| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Inferior function calls are powerful but might lead to undesired
results such as crashes when calling nested functions (frequently
used in particular in Ada).
This implements a GDB setting to disable calling inferior functions.
Note: the idea is that if/when the 'slash command' patch is pushed,
that this setting can be changed e.g. by using the shortcut /c.
This is version 2 of the patch. It handles all the received comments,
mostly replace 'can-call' by 'may-call', and avoid using
'inferior function call' in factor of 'calling function in the program'.
2019-04-26 Philippe Waroquiers <philippe.waroquiers@skynet.be>
gdb/ChangeLog
* NEWS: Mention the new set|show may-call-functions.
* infcall.c (may_call_functions_p): New variable.
(show_may_call_functions_p): New function.
(call_function_by_hand_dummy): Throws an error if not
may-call-functions.
(_initialize_infcall): Call add_setshow_boolean_cmd for
may-call-functions.
gdb/testsuite/ChangeLog
* gdb.base/callexit.exp: Test may-call-functions off.
gdb/doc/ChangeLog
* gdb.texinfo (Calling): Document the new
set|show may-call-functions.
|
|
This bug finds another usage where we end up segfaulting while
normalizing user input. inspect_type and replace_type recurse,
attempting to substitute the "real" symbol name for the typedef name.
However, since the both these names are the same, they keep calling
each other until the stack overflows.
A simple reproducer for it is given by
typedef struct foo foo;
int qux (foo *f) { return 0; }
(gdb) b qux(foo*)
Segmentation fault
inspect_type already contains some special handling to prevent a
similar situation from occurring with namespaces. I wonder, however,
whether we need be so pedantic about the exact nature of the substitution.
This patch implements this rather more aggressive assumption that these
substitutions should be avoided whenever the replacement symbol's name is
exactly the same as the one we're trying to substitute. [In the above
example, we're trying to substitute the tyepdef named "foo" with the symbol
named "foo" (a struct).]
gdb/ChangeLog:
PR c++/24367
* cp-support.c (inspect_type): Don't attempt substitutions
of symbol with the same name.
gdb/testsuite/ChangeLog:
PR c++/24367
* gdb.cp/meth-typedefs.cc (incomplete_struct)
(another_incomplete_struct, test_incomplete): New definitions.
(main): Use new definitions.
* gdb.cp/meth-typedefs.exp: Add new tests for `test_incomplete'
functions.
|
|
DW_FORM_strx is the new name of DW_FORM_GNU_str_index in the Dwarf 5 standard.
This is a small step towards supporting Dwarf 5 in gdb.
|
|
I forgot to include the ChangeLog entries in the commit
57e5e645010430b3d73f8c6a757d09f48dc8f8d5 ("Implement dump of mappings
with ELF headers by gcore").
|
|
This patch has a long story, but it all started back in 2015, with
commit df8411da087dc05481926f4c4a82deabc5bc3859 ("Implement support
for checking /proc/PID/coredump_filter"). The purpose of that commit
was to bring GDB's corefile generation closer to what the Linux kernel
does. However, back then, I did not implement the full support for
the dumping of memory mappings containing ELF headers (like mappings
of DSOs or executables). These mappings were being dumped most of
time, though, because the default value of /proc/PID/coredump_filter
is 0x33, which would cause anonymous private mappings (DSOs/executable
code mappings have this type) to be dumped. Well, until something
happened on binutils...
A while ago, I noticed something strange was happening with one of our
local testcases on Fedora GDB: it was failing due to some strange
build-id problem. On Fedora GDB, we (unfortunately) carry a bunch of
"local" patches, and some of these patches actually extend upstream's
build-id support in order to generate more useful information for the
user of a Fedora system (for example, when the user loads a corefile
into GDB, we detect whether the executable that generated that
corefile is present, and if it's not we issue a warning suggesting
that it should be installed, while also providing the build-id of the
executable). A while ago, Fedora GDB stopped printing those warnings.
I wanted to investigate this right away, and spent some time trying to
determine what was going on, but other things happened and I got
sidetracked. Meanwhile, the bug started to be noticed by some of our
users, and its priority started changing. Then, someone on IRC also
mentioned the problem, and when I tried helping him, I noticed he
wasn't running Fedora. Hm... So maybe the bug was *also* present
upstream.
After "some" time investigating, and with a lot of help from Keith and
others, I was finally able to determine that yes, the bug is also
present upstream, and that even though it started with a change in ld,
it is indeed a GDB issue.
So, as I said, the problem started with binutils, more specifically
after the following commit was pushed:
commit f6aec96dce1ddbd8961a3aa8a2925db2021719bb
Author: H.J. Lu <hjl.tools@gmail.com>
Date: Tue Feb 27 11:34:20 2018 -0800
ld: Add --enable-separate-code
This commit makes ld use "-z separate-code" by default on x86-64
machines. What this means is that code pages and data pages are now
separated in the binary, which is confusing GDB when it tries to decide
what to dump.
BTW, Fedora 28 binutils doesn't have this code, which means that
Fedora 28 GDB doesn't have the problem. From Fedora 29 on, binutils
was rebased and incorporated the commit above, which started causing
Fedora GDB to fail.
Anyway, the first thing I tried was to pass "-z max-page-size" and
specify a bigger page size (I saw a patch that did this and was
proposed to Linux, so I thought it might help). Obviously, this
didn't work, because the real "problem" is that ld will always use
separate pages for code and data. So I decided to look into how GDB
dumped the pages, and that's where I found the real issue.
What happens is that, because of "-z separate-code", the first two pages
of the ELF binary are (from /proc/PID/smaps):
00400000-00401000 r--p 00000000 fc:01 799548 /file
Size: 4 kB
KernelPageSize: 4 kB
MMUPageSize: 4 kB
Rss: 4 kB
Pss: 4 kB
Shared_Clean: 0 kB
Shared_Dirty: 0 kB
Private_Clean: 4 kB
Private_Dirty: 0 kB
Referenced: 4 kB
Anonymous: 0 kB
LazyFree: 0 kB
AnonHugePages: 0 kB
ShmemPmdMapped: 0 kB
Shared_Hugetlb: 0 kB
Private_Hugetlb: 0 kB
Swap: 0 kB
SwapPss: 0 kB
Locked: 0 kB
THPeligible: 0
VmFlags: rd mr mw me dw sd
00401000-00402000 r-xp 00001000 fc:01 799548 /file
Size: 4 kB
KernelPageSize: 4 kB
MMUPageSize: 4 kB
Rss: 4 kB
Pss: 4 kB
Shared_Clean: 0 kB
Shared_Dirty: 0 kB
Private_Clean: 0 kB
Private_Dirty: 4 kB
Referenced: 4 kB
Anonymous: 4 kB
LazyFree: 0 kB
AnonHugePages: 0 kB
ShmemPmdMapped: 0 kB
Shared_Hugetlb: 0 kB
Private_Hugetlb: 0 kB
Swap: 0 kB
SwapPss: 0 kB
Locked: 0 kB
THPeligible: 0
VmFlags: rd ex mr mw me dw sd
Whereas before, we had only one:
00400000-00401000 r-xp 00000000 fc:01 798593 /file
Size: 4 kB
KernelPageSize: 4 kB
MMUPageSize: 4 kB
Rss: 4 kB
Pss: 4 kB
Shared_Clean: 0 kB
Shared_Dirty: 0 kB
Private_Clean: 0 kB
Private_Dirty: 4 kB
Referenced: 4 kB
Anonymous: 4 kB
LazyFree: 0 kB
AnonHugePages: 0 kB
ShmemPmdMapped: 0 kB
Shared_Hugetlb: 0 kB
Private_Hugetlb: 0 kB
Swap: 0 kB
SwapPss: 0 kB
Locked: 0 kB
THPeligible: 0
VmFlags: rd ex mr mw me dw sd
Notice how we have "Anonymous" data mapped into the page. This will be
important.
So, the way GDB decides which pages it should dump has been revamped
by my patch in 2015, and now it takes the contents of
/proc/PID/coredump_filter into account. The default value for Linux
is 0x33, which means:
Dump anonymous private, anonymous shared, ELF headers and HugeTLB
private pages.
Or:
filter_flags filterflags = (COREFILTER_ANON_PRIVATE
| COREFILTER_ANON_SHARED
| COREFILTER_ELF_HEADERS
| COREFILTER_HUGETLB_PRIVATE);
Now, it is important to keep in mind that GDB doesn't always have *all*
of the necessary information to exactly determine the type of a page, so
the whole algorithm is based on heuristics (you can take a look at
linux-tdep.c:dump_mapping_p and
linux-tdep.c:linux_find_memory_regions_full for more info).
Before the patch to make ld use "-z separate-code", the (single) page
containing data and code was being flagged as an anonymous (due to the
non-zero "Anonymous:" field) private (due to the "r-xp" permission),
which means that it was being dumped into the corefile. That's why it
was working fine.
Now, as you can imagine, when "-z separate-code" is used, the *data*
page (which is where the ELF notes are, including the build-id one) now
doesn't have any "Anonymous:" mapping, so the heuristic is flagging it
as file-backed private, which is *not* dumped by default.
The next question I had to answer was: how come a corefile generated by
the Linux kernel was correct? Well, the answer is that GDB, unlike
Linux, doesn't actually implement the COREFILTER_ELF_HEADERS support.
On Linux, even though the data page is also treated as a file-backed
private mapping, it is also checked to see if there are any ELF headers
in the page, and then, because we *do* have ELF headers there, it is
dumped.
So, after more time trying to think of ways to fix this, I was able to
implement an algorithm that reads the first few bytes of the memory
mapping being processed, and checks to see if the ELF magic code is
present. This is basically what Linux does as well, except that, if
it finds the ELF magic code, it just dumps one page to the corefile,
whereas GDB will dump the whole mapping. But I don't think that's a
big issue, to be honest.
It's also important to explain that we *only* perform the ELF magic
code check if:
- The algorithm has decided *not* to dump the mapping so far, and;
- The mapping is private, and;
- The mapping's offset is zero, and;
- The user has requested us to dump mappings with ELF headers.
IOW, we're not going to blindly check every mapping.
As for the testcase, I struggled even more trying to write it. Since
our build-id support on upstream GDB is not very extensive, it's not
really possible to determine whether a corefile contains build-id
information or not just by using GDB. So, after thinking a lot about
the problem, I decided to rely on an external tool, eu-unstrip, in
order to verify whether the dump was successful. I verified the test
here on my machine, and everything seems to work as expected (i.e., it
fails without the patch, and works with the patch applied). We are
working hard to upstream our "local" Fedora GDB patches, and we intend
to submit our build-id extension patches "soon", so hopefully we'll be
able to use GDB itself to perform this verification.
I built and regtested this on the BuildBot, and no problems were
found.
gdb/ChangeLog:
2019-04-25 Sergio Durigan Junior <sergiodj@redhat.com>
PR corefiles/11608
PR corefiles/18187
* linux-tdep.c (dump_mapping_p): Add new parameters ADDR and
OFFSET. Verify if current mapping contains an ELF header.
(linux_find_memory_regions_full): Adjust call to
dump_mapping_p.
gdb/testsuite/ChangeLog:
2019-04-25 Sergio Durigan Junior <sergiodj@redhat.com>
PR corefiles/11608
PR corefiles/18187
* gdb.base/coredump-filter-build-id.exp: New file.
|
|
Add both board option and environment variable which enables gdbserver
debug and sends it to the file gdbserver.debug, located in the output
directory for the current test. Document this.
Add support for the environment variable in the Makefile.
The testsuite can be run with gdbserver debug enabled in the following way:
make check GDBSERVER_DEBUG=all
Disable tspeed.exp when debugging to prevent the log file filling
many gigabytes then timing out.
gdb/testsuite/ChangeLog:
* Makefile.in: Pass through GDBSERVER_DEBUG.
* README (Testsuite Parameters): Add GDBSERVER_DEBUG.
(gdbserver,debug): Add board setting.
* gdb.trace/tspeed.exp: Skip when debugging.
* lib/gdb.exp (gdbserver_debug_enabled): New procedure.
* lib/gdbserver-support.exp: Likewise
|
|
This changes the gdb test suite to omit -fno-stack-protector when
compiling Rust code. This makes Rust testing work again.
I think I saw this patch somewhere already, but I couldn't find it
again just now, so I'm checking this version in.
gdb/testsuite/ChangeLog
2019-04-24 Tom Tromey <tromey@adacore.com>
* lib/gdb.exp (gdb_compile): Don't add -fno-stack-protector for
Rust.
|
|
Commit 4aa866af ("Fix AMD64 return value ABI in expression
evaluation") introduced a regression when calling a function with a
structure that contains bitfields.
Because the caller of amd64_has_unaligned_fields handles bitfields
already, it seemed to me that the simplest fix was to ignore bitfields
here.
gdb/ChangeLog
2019-04-24 Tom Tromey <tromey@adacore.com>
* amd64-tdep.c (amd64_has_unaligned_fields): Ignore bitfields.
gdb/testsuite/ChangeLog
2019-04-24 Tom Tromey <tromey@adacore.com>
* gdb.arch/amd64-eval.exp: Test bitfield return.
* gdb.arch/amd64-eval.cc (struct Bitfields): New.
(class Foo) <return_bitfields>: New method.
(main): Call it.
|
|
Replaces use of aarch64_type_align with common type_align function.
Doing this fixes a bug in aarch64_type_align where static fields are
considered as part of the alignment calculation of a struct, which
results in arguments passed on the stack being misaligned. This bug
is exposed in the new test gdb.cp/many-args.exp.
Part of the old aarch64_type_align is retained and used as the gdbarch
type align callback in order to correctly align vectors.
gdb/ChangeLog:
* aarch64-tdep.c (aarch64_type_align): Only handle vector override
case.
(pass_on_stack): Use type_align.
(aarch64_gdbarch_init): Register aarch64_type_align gdbarch
function.
gdb/testsuite/ChangeLog:
* gdb.cp/many-args.cc: New file.
* gdb.cp/many-args.exp: New file.
|
|
When running gdb.btrace/reconnect.exp with native-gdbserver, we run into:
...
FAIL: gdb.btrace/reconnect.exp: first: stepi 19
...
due to the fact that we're trying to match:
...
stepi 19^M
0x00007ffff7dd8b57 in _dl_start () from /lib64/ld-linux-x86-64.so.2^M
...
using pattern:
...
gdb_test "stepi 19" "0x.* in .* from target.*"
...
Fix this by changing the pattern to:
...
gdb_test "stepi 19" "0x.* in .* from .*"
...
Tested on x86_64-linux with native and native-gdbserver.
gdb/testsuite/ChangeLog:
2019-04-23 Tom de Vries <tdevries@suse.de>
PR gdb/24433
* gdb.btrace/reconnect.exp: Fix stepi 19 pattern.
|
|
Ubuntu/Debian defaults PIE to enabled. This causes the trace tests
to fall over due to variables being returned as "unavailable". The
tests were never designed to work with pie.
Simply ensure the nopie flag is always used for the failing tests.
This removes 100+ failures when running native-gdbserver on Ubuntu 18.04.
gdb/testsuite/ChangeLog:
* gdb.trace/backtrace.exp: Use nopie flag.
* gdb.trace/circ.exp: Likewise.
* gdb.trace/collection.exp: Likewise.
* gdb.trace/ftrace.exp: Likewise.
* gdb.trace/mi-trace-unavailable.exp: Likewise.
* gdb.trace/mi-traceframe-changed.exp: Likewise.
* gdb.trace/qtro.exp: Likewise.
* gdb.trace/read-memory.exp: Likewise.
* gdb.trace/report.exp: Likewise.
* gdb.trace/tfile.exp: Likewise.
* gdb.trace/tfind.exp: Likewise.
* gdb.trace/unavailable.exp: Likewise.
|
|
On systems that use the probes-based solib interface, GDB misbehaves
if you run the "nosharelibrary" command, continue execution, and then
the program hits the shared library event breakpoint. On my system it
aborts like this:
(gdb) nosharedlibrary
(gdb) c
Continuing.
pure virtual method called
terminate called without an active exception
Aborted (core dumped)
Though it's really undefined behavior territory, caused by deferencing
a dangling solib event probe pointer.
I've observed this by running "nosharedlibrary" when stopped at the
entry point, but it should happen at any other point, if the program
does a dlopen/dlclose after.
The fix is to discard an objfile's probes from the svr4 probes table
when an objfile is about to be released.
New test included, works with both native and gdbserver testing.
Valgrind log:
(gdb) starti
(gdb) nosharedlibrary
(gdb) c
Continuing.
==24895== Invalid read of size 8
==24895== at 0x89E5FB: solib_event_probe_action(probe_and_action*) (solib-svr4.c:1735)
==24895== by 0x89E95A: svr4_handle_solib_event() (solib-svr4.c:1872)
==24895== by 0x8A7198: handle_solib_event() (solib.c:1274)
==24895== by 0x4E3407: bpstat_stop_status(address_space const*, unsigned long, thread_info*, target_waitstatus const*, bpstats*) (breakpoint.c:5407)
==24895== by 0x721F41: handle_signal_stop(execution_control_state*) (infrun.c:5685)
==24895== by 0x720B11: handle_inferior_event(execution_control_state*) (infrun.c:5129)
==24895== by 0x71DD93: fetch_inferior_event(void*) (infrun.c:3748)
==24895== by 0x7059C3: inferior_event_handler(inferior_event_type, void*) (inf-loop.c:43)
==24895== by 0x874DF0: remote_async_serial_handler(serial*, void*) (remote.c:14039)
==24895== by 0x894101: run_async_handler_and_reschedule(serial*) (ser-base.c:137)
==24895== by 0x8941E6: fd_event(int, void*) (ser-base.c:188)
==24895== by 0x67AFEF: handle_file_event(file_handler*, int) (event-loop.c:732)
==24895== Address 0x18b63860 is 0 bytes inside a block of size 136 free'd
==24895== at 0x4C2E616: operator delete(void*, unsigned long) (vg_replace_malloc.c:585)
==24895== by 0x8C6A12: stap_probe::~stap_probe() (stap-probe.c:124)
==24895== by 0x66F7DB: probe_key_free(bfd*, void*) (elfread.c:1382)
==24895== by 0x69B705: bfdregistry_callback_adaptor(void (*)(registry_container*, void*), registry_container*, void*) (gdb_bfd.c:131)
==24895== by 0x855A57: registry_clear_data(registry_data_registry*, void (*)(void (*)(registry_container*, void*), registry_container*, void*), registry_container*, registry_fields*) (registry.c:79)
==24895== by 0x855B01: registry_container_free_data(registry_data_registry*, void (*)(void (*)(registry_container*, void*), registry_container*, void*), registry_container*, registry_fields*) (registry.c:92)
==24895== by 0x69B783: bfd_free_data(bfd*) (gdb_bfd.c:131)
==24895== by 0x69C4BA: gdb_bfd_unref(bfd*) (gdb_bfd.c:609)
==24895== by 0x7CC33F: objfile::~objfile() (objfiles.c:651)
==24895== by 0x7CD559: objfile_purge_solibs() (objfiles.c:1021)
==24895== by 0x8A7132: no_shared_libraries(char const*, int) (solib.c:1252)
==24895== by 0x548E3D: do_const_cfunc(cmd_list_element*, char const*, int) (cli-decode.c:106)
==24895== Block was alloc'd at
==24895== at 0x4C2D42A: operator new(unsigned long) (vg_replace_malloc.c:334)
==24895== by 0x8C527C: handle_stap_probe(objfile*, sdt_note*, std::vector<probe*, std::allocator<probe*> >*, unsigned long) (stap-probe.c:1561)
==24895== by 0x8C5535: stap_static_probe_ops::get_probes(std::vector<probe*, std::allocator<probe*> >*, objfile*) const (stap-probe.c:1656)
==24895== by 0x66F71B: elf_get_probes(objfile*) (elfread.c:1365)
==24895== by 0x7EDD85: find_probes_in_objfile(objfile*, char const*, char const*) (probe.c:227)
==24895== by 0x4DF382: create_longjmp_master_breakpoint() (breakpoint.c:3275)
==24895== by 0x4F6562: breakpoint_re_set() (breakpoint.c:13828)
==24895== by 0x8A66AA: solib_add(char const*, int, int) (solib.c:1010)
==24895== by 0x89F7C6: enable_break(svr4_info*, int) (solib-svr4.c:2360)
==24895== by 0x8A104C: svr4_solib_create_inferior_hook(int) (solib-svr4.c:2992)
==24895== by 0x8A70B9: solib_create_inferior_hook(int) (solib.c:1215)
==24895== by 0x70C073: post_create_inferior(target_ops*, int) (infcmd.c:467)
==24895==
pure virtual method called
terminate called without an active exception
==24895==
==24895== Process terminating with default action of signal 6 (SIGABRT): dumping core
==24895== at 0x7CF3750: raise (raise.c:51)
==24895== by 0x7CF4D30: abort (abort.c:79)
==24895== by 0xB008F4: __gnu_cxx::__verbose_terminate_handler() (in build/gdb/gdb)
==24895== by 0xAFF845: __cxxabiv1::__terminate(void (*)()) (in build/gdb/gdb)
==24895== by 0xAFF890: std::terminate() (in build/gdb/gdb)
==24895== by 0xAFF95E: __cxa_pure_virtual (in build/gdb/gdb)
==24895== by 0x89E610: solib_event_probe_action(probe_and_action*) (solib-svr4.c:1735)
==24895== by 0x89E95A: svr4_handle_solib_event() (solib-svr4.c:1872)
==24895== by 0x8A7198: handle_solib_event() (solib.c:1274)
==24895== by 0x4E3407: bpstat_stop_status(address_space const*, unsigned long, thread_info*, target_waitstatus const*, bpstats*) (breakpoint.c:5407)
==24895== by 0x721F41: handle_signal_stop(execution_control_state*) (infrun.c:5685)
==24895== by 0x720B11: handle_inferior_event(execution_control_state*) (infrun.c:5129)
==24895==
Note, this little bit in the patch is just a cleanup that I noticed:
- lookup.prob = prob;
lookup.address = address;
That line isn't necessary because hashing/comparison only looks at the
address.
gdb/ChangeLog:
2019-04-22 Pedro Alves <palves@redhat.com>
* solib-svr4.c (svr4_free_objfile_observer): New.
(probe_and_action::objfile): New field.
(probes_table_htab_remove_objfile_probes)
(probes_table_remove_objfile_probes): New functions.
(register_solib_event_probe): Add 'objfile' parameter. Store it
in the new probe_and_action. Don't store the probe in 'lookup'.
(svr4_create_probe_breakpoints): Pass objfile to
register_solib_event_probe.
(_initialize_svr4_solib): Register a free_objfile observer.
gdb/testsuite/ChangeLog:
2019-04-22 Pedro Alves <palves@redhat.com>
* gdb.base/solib-probes-nosharedlibrary.c,
gdb.base/solib-probes-nosharedlibrary.exp: New files.
|
|
ada-lang.c is a bit too eager trying to decode unions in the Ada style
-- looking for discriminants and such. This causes crashes when
printing a non-Ada union in Ada mode, something that can easily happen
when printing a value from history or certain registers on AArch64.
This patch fixes the bug by changing ada-lang.c to only apply special
Ada treatment to types coming from an Ada CU. This in turn required a
couple of surprising changes.
First, some of the Ada code was already using HAVE_GNAT_AUX_INFO to
decide whether a type had already been fixed -- such types had
INIT_CPLUS_SPECIFIC called on them. This patch changes these spots to
use the "none" identifier instead.
This then required changing value_rtti_type to avoid changing the
language-specific object attached to an Ada type, which seems like a
good change regardless.
Tested on x86-64 Fedora 29.
gdb/ChangeLog
2019-04-19 Tom Tromey <tromey@adacore.com>
* ada-lang.c (ada_is_variant_part, ada_to_fixed_type_1):
Check ADA_TYPE_P.
(empty_record, ada_template_to_fixed_record_type_1)
(template_to_static_fixed_type)
(to_record_with_fixed_variant_part): Use INIT_NONE_SPECIFIC.
* cp-abi.c (value_rtti_type): Check HAVE_CPLUS_STRUCT.
* gdbtypes.h (INIT_NONE_SPECIFIC, ADA_TYPE_P): New
macros.
gdb/testsuite/ChangeLog
2019-04-19 Tom Tromey <tromey@adacore.com>
* gdb.ada/ptype_union.c: New file.
* gdb.ada/ptype_union.exp: New file.
|
|
PR symtab/24423 points out that control characters in a source file
cause a hang in the "list" command, a regression introduced by the
styling changes.
This patch, from the PR, fixes the bug. I've included a minimal
change to the "list" test that exercises this code.
I recall that this bug was discussed on gdb-patches, and I thought
there was a patch there as well, but I was unable to find it.
gdb/ChangeLog
2019-04-19 Ilya Yu. Malakhov <malakhov@mcst.ru>
PR symtab/24423:
* source.c (print_source_lines_base): Advance "iter" when a
control character is seen.
gdb/testsuite/ChangeLog
2019-04-19 Tom Tromey <tromey@adacore.com>
PR symtab/24423:
* gdb.base/list0.h (foo): Add a control-l character.
|
|
When running break-probes.exp with native-gdbserver, we run into:
...
FAIL: gdb.base/break-probes.exp: run til our library loads (the program exited)
FAIL: gdb.base/break-probes.exp: call (int) foo(23)
...
due to the fact that we're trying to match:
...
Inferior loaded /data/gdb_versions/devel/build/gdb/testsuite/outputs/gdb.base\
/break-probes/break-probes-solib.so
...
using pattern:
...
Inferior loaded $sysroot$binfile_lib
...
which expands into:
...
Inferior loaded //data/gdb_versions/devel/build/gdb/testsuite/outputs/gdb.base\
/break-probes/break-probes-solib.so
...
Fix by setting sysroot to "" in local-board.exp.
Tested on x86_64-linux with native-gdbserver.
gdb/testsuite/ChangeLog:
2019-04-18 Tom de Vries <tdevries@suse.de>
PR gdb/24433
* boards/local-board.exp: Set sysroot to "".
|
|
When debugging any of the testcases added by this commit, which do a
vfork in a thread with "set follow-fork-mode child" + "set
detach-on-fork on", we run into this assertion:
...
src/gdb/nat/x86-linux-dregs.c:146: internal-error: \
void x86_linux_update_debug_registers(lwp_info*): \
Assertion `lwp_is_stopped (lwp)' failed.
...
The assert is caused by the following: the vfork-child exit or exec
event is handled by handle_vfork_child_exec_or_exit, which calls
target_detach to detach from the vfork parent. During target_detach
we call linux_nat_target::detach, which:
#1 - stops all the threads
#2 - waits for all the threads to be stopped
#3 - detaches all the threads
However, during the second step we run into this code in
stop_wait_callback:
...
/* If this is a vfork parent, bail out, it is not going to report
any SIGSTOP until the vfork is done with. */
if (inf->vfork_child != NULL)
return 0;
...
and we don't wait for the threads to be stopped, which results in this
assert in x86_linux_update_debug_registers triggering during the third
step:
...
gdb_assert (lwp_is_stopped (lwp));
...
The fix is to reset the vfork parent's vfork_child field before
calling target_detach in handle_vfork_child_exec_or_exit. There's
already similar code for the other paths handled by
handle_vfork_child_exec_or_exit, so this commit refactors the code a
bit so that all paths share the same code.
The new tests cover both a vfork child exiting, and a vfork child
execing, since both cases would trigger the assertion.
The new testcases also exercise following the vfork children with "set
detach-on-fork off", since it doesn't seem to be tested anywhere.
Tested on x86_64-linux, using native and native-gdbserver.
gdb/ChangeLog:
2019-04-18 Tom de Vries <tdevries@suse.de>
Pedro Alves <palves@redhat.com>
PR gdb/24454
* infrun.c (handle_vfork_child_exec_or_exit): Reset vfork parent's
vfork_child field before calling target_detach.
gdb/testsuite/ChangeLog:
2019-04-18 Tom de Vries <tdevries@suse.de>
Pedro Alves <palves@redhat.com>
PR gdb/24454
* gdb.threads/vfork-follow-child-exec.c: New file.
* gdb.threads/vfork-follow-child-exec.exp: New file.
* gdb.threads/vfork-follow-child-exit.c: New file.
* gdb.threads/vfork-follow-child-exit.exp: New file.
|
|
The AMD64 System V ABI specifies that when a function has a return type
classified as MEMORY, the caller provides space for the value and passes
the address to this space as the first argument to the function (before
even the "this" pointer). The classification of MEMORY is applied to
struct that are sufficiently large, or ones with unaligned fields.
The expression evaluator uses call_function_by_hand to call functions,
and the hand-built frame has to push arguments in a way that matches the
ABI of the called function. call_function_by_hand supports ABI-based
struct returns, based on the value of gdbarch_return_value, however on
AMD64 the implementation of the classifier incorrectly assumed that all
non-POD types (implemented as "all types with a base class") should be
classified as MEMORY and use the struct return.
This ABI mismatch resulted in issues when calling a function that returns
a class of size <16 bytes which has a base class, including issues such
as the "this" pointer being incorrect (as it was passed as the second
argument rather than the first).
This is now fixed by checking for field alignment rather than POD-ness,
and a testsuite is added to test expression evaluation for AMD64.
gdb/ChangeLog:
* amd64-tdep.c (amd64_classify_aggregate): Use cp_pass_by_reference
rather than a hand-rolled POD check when checking for forced MEMORY
classification.
gdb/testsuite/ChangeLog:
* gdb.arch/amd64-eval.cc: New file.
* gdb.arch/amd64-eval.exp: New file.
|
|
The local board file ensures that the sysroot is always set to load
files from the local filesystem.
Add a gdbserver test to explicitly test the sysroot set to both the
remote target and the local filesystem.
gdb/testsuite/ChangeLog:
* gdb.server/sysroot.c: New test.
* gdb.server/sysroot.exp: New file.
* lib/gdbserver-support.exp (gdb_target_cmd): Add additional text
matching param.
|
|
The current code in gdbtypes.c:type_align incorrectly returns 0 as the
alignment for a structure containing only static fields. After this
patch the correct value of 1 is returned. The gdb.base/align.exp test
is extended to cover this case.
gdb/ChangeLog:
* gdbtypes.c (type_align): A struct with no non-static fields also
has alignment of 1.
gdb/testsuite/ChangeLog:
* gdb.base/align.exp: Extend test to cover structures containing
only static fields.
|
|
We can use CC_WITH_TWEAKS_FLAGS when cd-ing into the gdb build subdir and
invoking make check:
...
$ cd $objdir/gdb
$ make check \
RUNTESTFLAGS='--target_board=cc-with-tweaks' \
CC_WITH_TWEAKS_FLAGS='-z'
...
But when cd-ing into the top-level build dir and invoking make check-gdb
instead:
...
$ cd $objdir
$ make check-gdb \
RUNTESTFLAGS='--target_board=cc-with-tweaks' \
CC_WITH_TWEAKS_FLAGS='-z'
...
using CC_WITH_TWEAKS_FLAGS has no effect, because CC_WITH_TWEAKS_FLAGS is not
passed down from the top level Makefile.
Add cc-with-dwz.exp and cc-with-dwz-m.exp, that don't require
CC_WITH_TWEAKS_FLAGS to be set in the make invocation, allowing us to run these
test configurations from the toplevel build dir:
...
$ cd $objdir
$ make check-gdb \
RUNTESTFLAGS='--target_board=cc-with-dwz'
...
Tested on x86_64-linux.
gdb/testsuite/ChangeLog:
2019-04-11 Tom de Vries <tdevries@suse.de>
* boards/cc-with-dwz-m.exp: New file.
* boards/cc-with-dwz.exp: New file.
* boards/cc-with-tweaks.exp: Note that check-gdb doesn't work.
|
|
When using the "start" command, GDB puts a temporary breakpoint on the
"main" symbol (we literally invoke the tbreak command). However, since
it does wild matching by default, it also puts a breakpoint on any C++
method or "main" function in a namespace. For example, when debugging
GDB, it creates a total of 24 locations:
(gdb) start
Temporary breakpoint 1 at 0x198c1e9: main. (24 locations)
as there are a bunch of methods called main in the selftests, such as
selftests::string_view::capacity_1::main()
If such method was called in the constructor of a global object, or a
function marked with the attribute "constructor", then we would stop at
the wrong place. Also, this causes a few extra symtabs (those that
contain the "wrong" mains) to be expanded for nothing.
The dummiest, most straightforward solution is to add -qualified when
invoking tbreak. With this patch, "start" creates a single-location
breakpoint, as expected.
I copied the start.exp test to start-cpp.exp and made it use a C++ test
file, which contains two main functions. The new test verifies that the
output of "start" is the output we get when we set a single-location
breakpoint.
gdb/ChangeLog:
* infcmd.c (run_command_1): Pass -qualified to tbreak when usind
the "start" command.
gdb/testsuite/ChangeLog:
* gdb.base/start-cpp.exp: New file.
* gdb.base/start-cpp.cc: New file.
|
|
This renaming was done to stay consistent with the naming of the new
gdb.InferiorThread.handle method. I had initially named it "thread_handle"
but Tom Tromey suggested just "handle".
The old name (thread_from_thread_handle) still works, but is marked as
deprecated in comments in the code as well as in the documentation.
I have some code which uses these functions. I very much like the
brevity of the new names.
gdb/doc/ChangeLog:
* python.texi (Inferiors In Python): Rename
Inferior.thread_from_thread_handle to Inferior.thread_from_handle.
Add note about the former being deprecated.
gdb/ChangeLog:
* python/py-inferior.c (infpy_thread_from_thread_handle):
Adjust comments to reflect renaming of thread_from_thread_handle
to thread_from_handle. Adjust keywords. Fix type error message.
(inferior_object_methods): Add thread_from_handle. Retain
thread_from_thread_handle, but mark it as deprecated.
testsuite/ChangeLog:
* gdb.python/py-thrhandle.exp: Adjust tests to call
thread_from_handle instead of thread_from_thread_handle.
|
|
gdb/testsuite/ChangeLog:
* gdb.python/py-thrhandle.exp: Add tests for
gdb.InferiorThread.handle.
|
|
If an convenience function is defined in python (or guile), then
currently this will not work in Fortran, instead the user is given
this message:
(gdb) set language fortran
(gdb) p $myfunc (3)
Cannot perform substring on this type
Compare this to C:
(gdb) set language c
(gdb) p $myfunc (3)
$1 = 1
After this patch we see the same behaviour in both C and Fortran.
I've extended the test to check that all languages can call the
convenience functions - only Fortran was broken.
When calling convenience functions in Fortran we don't need to perform
the same value preparation (passing by pointer) that we would for
calling a native function - passing the real value is fine.
gdb/ChangeLog:
* eval.c (evaluate_subexp_standard): Handle internal functions
during Fortran function call handling.
gdb/testsuite/ChangeLog:
* gdb.python/py-function.exp: Check calling helper function from
all languages.
* lib/gdb.exp (gdb_supported_languages): New proc.
|
|
Add two new internal functions $_cimag and $_creal that extract the
imaginary and real parts of a complex value.
These internal functions can take a complex value of any type 'float
complex', 'double complex', or 'long double complex' and return a
suitable floating point value 'float', 'double', or 'long double'.
So we can now do this:
(gdb) p z1
$1 = 1.5 + 4.5 * I
(gdb) p $_cimag (z1)
$4 = 4.5
(gdb) p $_creal (z1)
$4 = 1.5
The components of a complex value are not strictly named types in
DWARF, as the complex type is itself the base type. However, once we
are able to extract the components it makes sense to be able to ask
what the type of these components is and get a sensible answer back,
rather than the error we would currently get. Currently GDB says:
(gdb) ptype z1
type = complex double
(gdb) p $_cimag (z1)
$4 = 4.5
(gdb) ptype $
type = <invalid type code 9>
With the changes in dwarf2read.c, GDB now says:
(gdb) ptype z1
type = complex double
(gdb) p $_cimag (z1)
$4 = 4.5
(gdb) ptype $
type = double
Which seems to make more sense.
gdb/ChangeLog:
* NEWS: Mention new internal functions.
* dwarf2read.c (dwarf2_init_complex_target_type): New function.
(read_base_type): Use dwarf2_init_complex_target_type.
* value.c (creal_internal_fn): New function.
(cimag_internal_fn): New function.
(_initialize_values): Register new internal functions.
gdb/doc/ChangeLog:
* gdb.texinfo (Convenience Funs): Document '$_creal' and
'$_cimag'.
gdb/testsuite/ChangeLog:
* gdb.base/complex-parts.c: New file.
* gdb.base/complex-parts.exp: New file.
|
|
add_partial_subprogram does not handle DW_AT_ranges, while the full
symtab reader does. This can lead to discrepancies where a function
is not put into a partial symtab, and so is not available to "break"
and the like -- but is available if the full symtab has somehow been
read.
This patch fixes the bug by arranging to read DW_AT_ranges when
reading partial DIEs.
This is PR symtab/23331.
The new test case is derived from dw2-ranges-func.exp, which is why I
kept the copyright dates.
gdb/ChangeLog
2019-04-01 Tom Tromey <tromey@adacore.com>
PR symtab/23331:
* dwarf2read.c (partial_die_info::read): Handle DW_AT_ranges.
gdb/testsuite/ChangeLog
2019-04-01 Tom Tromey <tromey@adacore.com>
PR symtab/23331:
* gdb.dwarf2/dw2-ranges-main.c: New file.
* gdb.dwarf2/dw2-ranges-psym.c: New file.
* gdb.dwarf2/dw2-ranges-psym.exp: New file.
|
|
The str () function, called on a gdb.Value instance, produces a string
representation similar to what can be achieved with the print command,
but it doesn't allow to specify additional formatting settings, for
instance disabling pretty printers.
This patch introduces a new format_string () method to gdb.Value which
allows specifying more formatting options, thus giving access to more
features provided by the internal C function common_val_print ().
gdb/ChangeLog:
2019-04-01 Marco Barisione <mbarisione@undo.io>
Add gdb.Value.format_string ().
* python/py-value.c (copy_py_bool_obj):
(valpy_format_string): Add gdb.Value.format_string ().
* NEWS: Document the addition of gdb.Value.format_string ().
gdb/doc/ChangeLog:
2019-04-01 Marco Barisione <mbarisione@undo.io>
* python.texi (Values From Inferior): Document
gdb.Value.format_string ().
gdb/testsuite/ChangeLog:
2019-04-01 Marco Barisione <mbarisione@undo.io>
Test gdb.Value.format_string ().
* gdb.python/py-format-string.exp: New test.
* gdb.python/py-format-string.c: New file.
* gdb.python/py-format-string.py: New file.
|
|
gdb/ChangeLog:
2019-03-30 Eli Zaretskii <eliz@gnu.org>
* NEWS: Announce $_gdb_major and $_gdb_minor.
* top.c (init_gdb_version_vars): New function.
(gdb_init): Call init_gdb_version_vars.
gdb/testsuite/ChangeLog:
2019-03-30 Simon Marchi <simark@simark.ca>
* gdb.base/default.exp: Add values for $_gdb_major and
$_gdb_minor.
gdb/doc/ChangeLog:
2019-03-30 Eli Zaretskii <eliz@gnu.org>
* gdb.texinfo (Convenience Vars): Document $_gdb_major and
$_gdb_minor.
|
|
I noticed that the help for "info addr" did not include a "usage"
line; and when adding it I went through and fixed a few minor issues
in printcmd.c:
* Added usage lines to all commands
* Updated the help text for some commands
* Changed some help to use upper case metasyntactic variables
* Removed some dead code
Regression tested on x86-64 Fedora 29.
gdb/ChangeLog
2019-03-29 Tom Tromey <tromey@adacore.com>
* printcmd.c (_initialize_printcmd): Add usage lines. Update some
help text. Remove dead code.
gdb/testsuite/ChangeLog
2019-03-29 Tom Tromey <tromey@adacore.com>
* gdb.base/help.exp: Tighten apropos regexp.
|
|
This is the fortran part of the patch, including tests, which
are essentially unchanged from Siddhesh's original 2012 submission:
https://sourceware.org/ml/gdb-patches/2012-08/msg00562.html
There is, however, one large departure. In the above thread,
Jan pointed out problems with GCC debuginfo for -m32 builds
(filed usptream as gcc/54934). After investigating the issue,
I am dropping the hand-tweaked assembler source file to workaround
this case.
While I would normally do something to accommodate this, in
this case, given the ubiquity of 64-bit systems today (where
the tests pass) and the apparent lack of urgency on the compiler
side (by users), I don't think the additional complexity and
maintenance costs are worth it. It will be very routinely tested
on 64-bit systems. [For example, at Red Hat, we always
test -m64 and -m32 configurations for all GDB releases.]
gdb/ChangeLog:
From Siddhesh Poyarekar:
* f-lang.h (f77_get_upperbound): Return LONGEST.
(f77_get_lowerbound): Likewise.
* f-typeprint.c (f_type_print_varspec_suffix): Expand
UPPER_BOUND and LOWER_BOUND to LONGEST. Use plongest to format
print them.
(f_type_print_base): Expand UPPER_BOUND to LONGEST. Use
plongest to format print it.
* f-valprint.c (f77_get_lowerbound): Return LONGEST.
(f77_get_upperbound): Likewise.
(f77_get_dynamic_length_of_aggregate): Expand UPPER_BOUND,
LOWER_BOUND to LONGEST.
(f77_create_arrayprint_offset_tbl): Likewise.
gdb/testsuite/ChangeLog:
* gdb.fortran/array-bounds.exp: New file.
* gdb.fortran/array-bounds.f90: New file.
|
|
Similarly to multi-arch-exec.exp, increase the alarm timer to avoid
test blocking under high load or with a slow gdb.
2019-03-28 Philippe Waroquiers <philippe.waroquiers@skynet.be>
* gdb.multi/multi-term-settings.c (main): Increase alarm timer.
|
|
When running multi-arch-exec.exp under valgrind, the test succeeds
when the machine is not loaded, but blocks when the machine is highly
loaded (e.g. when running the testsuite with valgrind with -j X
where X is one more than the nr of available cores).
The problem is that the hello program dies too early due to the alarm (30).
So, increase the alarm timer.
Note that this does not make the test take longer (it takes about
3.5 seconds on my system). As I understand, the alarm is just there
to avoid hello staying there forever in case of another problem.
2019-03-28 Philippe Waroquiers <philippe.waroquiers@skynet.be>
* gdb.multi/hello.c (main): Increase alarm timer.
|
|
This patch fixes a problem on nios2-linux-gnu with stepping past the
kernel helper __kuser_cmpxchg, which was exposed by the testcase
gdb.threads/watchpoint-fork.exp. The kernel maps this function into
user space on an unwritable page. In this testcase, the cmpxchg
helper is invoked indirectly from the setbuf call in the test program.
Since this target lacks hardware breakpoint/watchpoint support, GDB
tries to single-step through the program by setting software
breakpoints, and was just giving an error when it reached the function
on the unwritable page.
The solution here is to always step over the call instead of stepping
into it; cmpxchg is supposed to be an atomic operation so this
behavior seems reasonable. The hook in nios2_get_next_pc is somewhat
generic, but at present cmpxchg is the only helper provided by the
Linux kernel that is invoked by an ordinary function call. (Signal
return trampolines also go through the unwritable page but not by a
function call.)
Fixing this issue also revealed that the testcase needs a much larger
timeout factor when software single-stepping is used. That has also
been fixed in this patch.
gdb/ChangeLog
2019-03-28 Sandra Loosemore <sandra@codesourcery.com>
* nios2-tdep.h (struct gdbarch_tdep): Add is_kernel_helper.
* nios2-tdep.c (nios2_get_next_pc): Skip over kernel helpers.
* nios2-linux-tdep.c (nios2_linux_is_kernel_helper): New.
(nios2_linux_init_abi): Install it.
gdb/testsuite/ChangeLog
2019-03-28 Sandra Loosemore <sandra@codesourcery.com>
* gdb.threads/watchpoint-fork.exp (test): Use large timeout
factor when no hardware watchpoint support.
|
|
When testing using native-gdbserver and native-extended-gdbserver, the sysroot
is not set. This results in a warning from GDB and files are sent via the
remote protocol, which can be slow.
On Ubuntu 18.04 (unlike most distros) the debug versions of the standard
libraries are included by default in /usr/lib/debug/.
These file reads are causing a complete native-gdbserver run on the AArch64
buildbot slave to timeout after 2.5 hours. This is also causing the builds
to back up on the slave.
The solution is to ensure the sysroot is set to / for all local boards.
This drastically reduces the time of a test. For example, gdb.base/sigall.exp
drops from 23 seconds to 4 seconds.
A full native-gdbserver run on the AArch64 slave now takes 8 minutes.
gdb/testsuite/ChangeLog:
* boards/local-board.exp: set sysroot to /.
|
|
Looking at the AArch64 buildbot, I noticed about two dozen old instances of
interrupt-daemon-attach taking up a full 100% cpu each.
If the test fails then the test binary relies on an alarm to ensure it dies
after 60 seconds.
As per the Linux man page for alarm:
Alarms created by alarm() ... are not inherited by children created via fork.
Update the test to add an alarm in the child and also put a sleep in the
child loop so it does not constantly consume cpu.
Note I haven't managed to re-create why the test failed. This fix will just
stop it hanging and consuming cpu when it does.
gdb/testsuite/ChangeLog:
* gdb.base/interrupt-daemon-attach.c (main): Add alarm and sleep
in child.
|
|
The documentation say that the display_hint method must return a
string to serve as a display hint, and then goes on to list some
acceptable strings.
However, if we don't supply the display_hint method then we get a
default display style behaviour and there's currently no way (in the
python api) to force this default behaviour.
The guile api allows #f to be used in order to force the default
display style behaviour, and this is documented.
Currently, using None in the python api also forces the default
display behaviour.
This commit extends the documentation to make returning None from the
display_hint method an official mechanism by which the user can get
the default display style.
I've extended one of the existing tests to cover this case.
gdb/doc/ChangeLog:
* python.texi (Pretty Printing API): Document use of None for the
display_hint.
gdb/testsuite/ChangeLog:
* gdb.python/py-prettyprint.c (struct container) <is_map_p>: New
field.
(make_container): Initialise new field.
* gdb.python/py-prettyprint.exp: Add new tests.
* gdb.python/py-prettyprint.py (class ContainerPrinter)
<display_hint>: New method.
|
|
This makes the test names unique in gdb.python/py-prettyprint.exp, it
also switches to use gdb_breakpoint and gdb_continue_to_breakpoint
more so that we avoid test names with the source line number in - this
is bad if the test source ever changes as the test names will then
change.
One final change is to switch from using gdb_py_test_silent_cmd to use
gdb_test_no_output, the former should be used for running python
commands and can catch any thrown exception. However, in this case
the command being run is not a python command, its just a normal GDB
CLI command that produces no output, so lets use the appropriate
wrapper function.
gdb/testsuite/ChangeLog:
* gdb.python/py-prettyprint.exp: Use gdb_breakpoint and
gdb_continue_to_breakpoint more throughout this test.
(run_lang_tests) Supply unique test names, and use
gdb_test_no_output.
|
|
While writing a new test for 'set print pretty on' I spotted that GDB
will sometimes add a trailing whitespace character when pretty
printing. This commit removes the trailing whitespace and updates the
expected results in one tests where this was an issue.
I've added an extra test for 'set print pretty on' as it doesn't seem
to have much testing.
gdb/ChangeLog:
* cp-valprint.c (cp_print_value_fields): Don't print trailing
whitespace when pretty printing is on.
gdb/testsuite/ChangeLog:
* gdb.base/finish-pretty.exp: Update expected results.
* gdb.base/pretty-print.c: New file.
* gdb.base/pretty-print.exp: New file.
|
|
This commit fixes a regression in the testsuite itself, triggered by
errors being raised from within gdb_test_multiple, originally reported
by Pedro Franco de Carvalho's at
<https://sourceware.org/ml/gdb-patches/2019-03/msg00160.html>. Parts
of the commit message are based on his report.
This started happening due to a commit that was introduced recently,
and it can cause the testsuite to hang.
The commit that triggers this is:
fe1a5cad302b5535030cdf62895e79512713d738
[gdb/testsuite] Log wait status on process no longer exists error
That commit introduces a new "eof" block in gdb_test_multiple. That
is not incorrect itself, but dejagnu's remote_expect is picking that
block as the "default" action when an error is raised from within the
commands inside a call to gdb_test_multiple:
# remote_expect works basically the same as standard expect, but it
# also takes care of getting the file descriptor from the specified
# host and also calling the timeout/eof/default section if there is an
# error on the expect call.
#
proc remote_expect { board timeout args } {
I find that "feature" surprising, and I don't really know why it
exists, but this means that the eof section that remote_expect picks
as the error block can be executed even when there was no actual eof
and the GDB process is still running, so the wait introduced in the
commit that tries to get the exit status of GDB hangs forever, while
GDB itself waits for input.
This only happens when there are internal testsuite errors (not
testcase failures). This can be reproduced easily with a testcase
such as:
gdb_start
gdb_test_multiple "show version" "show version" {
-re ".*" {
error "forced error"
}
}
I think that working around this in GDB is useful so that the
testsuite doesn't hang in these cases.
Adding an empty "default" block at the end of the expect body in
gdb_test_multiple doesn't work, because dejagnu gives preference to
"eof" blocks:
if { $x eq "eof" } {
set save_next 1
} elseif { $x eq "default" || $x eq "timeout" } {
if { $error_sect eq "" } {
set save_next 1
}
}
And we do have "eof" blocks. So we need to make sure that the last
"eof" block is safe to use as the default error block. It's also
pedantically incorrect to print
"ERROR: Process no longer exists"
which is what we'd get if the last eof block we have was selected
(more below on this).
So this commit solves this by appending an "eof" with an empty
spawn_id list, so that it won't ever match.
Now, why is the first "eof" block selected today as the error block,
instead of the last one?
The reason is that remote_expect, while parsing the body to select the
default block to execute after an error, is affected by the comments
in the body (since they are also parsed).
If this comment in gdb_test_multiple
# patterns below apply to any spawn id specified.
is changed to
# The patterns below apply to any spawn id specified.
then the second eof block is selected and there is no hang.
Any comment at that same place with an even number of tokens also
works.
This is IMO a coincidence caused by how comments work in TCL.
Comments should only appear in places where a command can appear. And
here, remote_expect is parsing a list of options, not commands, so
it's not unreasonable to not parse comments, similarly to how this:
set a_list {
an_element
# another_element
}
results in a list with three elements, not one element.
The fact that comments with an even number of tokens work is just a
coincidence of how remote_expect's little state machine is
implemented.
I thought we could solve this by stripping out comment lines in
gdb_expect, but I didn't find an easy way to do that. Particularly, a
couple naive approaches I tried run into complications. For example,
we have gdb_test calls with regular expressions that include sequences
like "\r\n#", and by the time we get to gdb_expect, the \r\n have
already been expanded to a real newline, so just splitting the whole
body at newline boundaries, looking for lines that start with #
results in incorrectly stripping out half of the gdb_text regexp. I
think it's better (at least in this commit), to move the comments out
of the list, because it's much simpler and risk free.
gdb/testsuite/ChangeLog:
2019-03-25 Pedro Alves <palves@redhat.com>
* lib/gdb.exp (gdb_test_multiple): Split appends to $code and
move comments outside list. Append '-i "" eof' section.
|
|
Recent versions of Ubuntu and Debian default GCC to enable pie.
In dump.exp, pie will causes addresses to be out of range for IHEX.
In break-interp.exp, pie is explicitly set for some tests and assumed
to be disabled for the remainder.
Ensure pie is disabled for these tests when required.
In addition, add a pie option to gdb_compile to match the nopie option
and simplify use.
gdb/testsuite/ChangeLog:
* README: Add pie options.
* gdb.base/break-interp.exp: Ensure pie is disabled.
* gdb.base/dump.exp: Likewise.
* lib/gdb.exp (gdb_compile): Add pie option.
|
|
If you run "gdb -i=mi2" and set a "display", then when "next"ing the
displays will be shown twice:
~"1: x = 23\n"
~"7\t printf(\"%d\\n\", x);\n"
~"1: x = 23\n"
*stopped,reason="end-stepping-range",frame={addr="0x0000000000400565",func="main",args=[],file="q.c",fullname="/tmp/q.c",line="7"},thread-id="1",stopped-threads="all",core="1"
The immediate cause of this is this code in mi_on_normal_stop_1:
print_stop_event (mi_uiout);
console_interp = interp_lookup (current_ui, INTERP_CONSOLE);
if (should_print_stop_to_console (console_interp, tp))
print_stop_event (mi->cli_uiout);
... which obviously prints the stop twice.
However, I think the first call to print_stop_event is intended just
to emit the MI *stopped notification, which explains why the source
line does not show up two times.
This patch fixes the bug by changing print_stop_event to only call
do_displays for non-MI-like ui-outs.
Tested on x86-64 Fedora 29.
gdb/ChangeLog
2019-03-19 Tom Tromey <tromey@adacore.com>
* mi/mi-interp.c (mi_on_normal_stop_1): Only show displays once.
* infrun.h (print_stop_event): Add "displays" parameter.
* infrun.c (print_stop_event): Add "displays" parameter.
gdb/testsuite/ChangeLog
2019-03-19 Tom Tromey <tromey@adacore.com>
* gdb.mi/mi2-cli-display.c: New file.
* gdb.mi/mi2-cli-display.exp: New file.
|
|
Using ptype on an array type in Ada can sometimes show an incorrect
high bound. This happens because ada_evaluate_subexp will create an
array with an incorrect upper bound in the EVAL_AVOID_SIDE_EFFECTS
case.
This patch fixes the problem by arranging to always create such an
array with valid bounds.
Tested on x86-64 Fedora 29.
gdb/ChangeLog
2019-03-18 Tom Tromey <tromey@adacore.com>
* ada-lang.c (empty_array): Add "high" parameter.
(ada_evaluate_subexp): Update.
gdb/testsuite/ChangeLog
2019-03-18 Joel Brobecker <brobecker@adacore.com>
Tom Tromey <tromey@adacore.com>
* gdb.ada/ptype_array/pck.adb: New file.
* gdb.ada/ptype_array/pck.ads: New file.
* gdb.ada/ptype_array/foo.adb: New file.
* gdb.ada/ptype_array.exp: New file.
|
|
This adds "set style source" (and "show style source") commands. This
gives the user control over whether source code is highlighted.
gdb/ChangeLog
2019-03-14 Tom Tromey <tromey@adacore.com>
* NEWS: Add item for "style sources" commands.
* source-cache.c (source_cache::get_source_lines): Check
source_styling.
* cli/cli-style.c (source_styling): New global.
(_initialize_cli_style): Add "style sources" commands.
(show_style_sources): New function.
* cli/cli-style.h (source_styling): Declare.
gdb/doc/ChangeLog
2019-03-14 Tom Tromey <tromey@adacore.com>
* gdb.texinfo (Output Styling): Document "set style source" and
"show style source".
gdb/testsuite/ChangeLog
2019-03-14 Tom Tromey <tromey@adacore.com>
* gdb.base/style.exp: Add "set style sources" test.
|
|
New in v2:
- Addressed comments about doc, updated the MI version table
- New doc for the Breakpoint information format
- New -fix-multi-location-breakpoint-output command, with associated
doc, test and NEWS updated accordingly
- Fixed the output, the locations list is now actually in the tuple
representing the breakpoint.
Various MI commands or events related to breakpoints output invalid MI
records when printing information about a multi-location breakpoint.
For example:
-break-insert allo
^done,bkpt={...,addr="<MULTIPLE>",...},{number="1.1",...},{number="1.2",...}
The problem is that according to the syntax [1], the top-level elements
are of type "result" and should be of the form "variable=value".
This patch changes the output to wrap the locations in a list:
^done,bkpt={...,addr="<MULTIPLE>",locations=[{number="1.1",...},{number="1.2",...}]}
The events =breakpoint-created, =breakpoint-modified, as well as the
-break-info command also suffer from this (and maybe others I didn't
find).
Since this is a breaking change for MI, we have to deal somehow with
backwards compatibility. The approach taken by this patch is to bump
the MI version, use the new syntax in MI3 while retaining the old syntax
in MI2. Frontends are expected to use a precise MI version (-i=mi2), so
if they do that they should be unaffected.
The patch also adds the command -fix-multi-location-breakpoint-output,
which front ends can use to enable this behavior with MI <= 2.
[1] https://sourceware.org/gdb/onlinedocs/gdb/GDB_002fMI-Output-Syntax.html#GDB_002fMI-Output-Syntax
gdb/ChangeLog:
* NEWS: Mention that the new default MI version is 3. Mention
changes to the output of commands and events that deal with
multi-location breakpoints.
* breakpoint.c: Include "mi/mi-out.h".
(print_one_breakpoint): Change output syntax if using MI version
>= 3.
* mi/mi-main.h (mi_cmd_fix_multi_location_breakpoint_output):
New.
(mi_multi_location_breakpoint_output_fixed): New.
* mi/mi-main.c (fix_multi_location_breakpoint_output): New.
(mi_cmd_fix_multi_location_breakpoint_output): New.
(mi_multi_location_breakpoint_output_fixed): New.
* mi/mi-cmds.c (mi_cmds): Register command
-fix-multi-location-breakpoint-output.
* mi/mi-out.c (mi_out_new): Instantiate version 3 when using
interpreter "mi".
gdb/testsuite/ChangeLog:
* mi-breakpoint-location-ena-dis.exp: Rename to ...
* mi-breakpoint-multiple-locations.exp: ... this.
(make_breakpoints_pattern): New proc.
(do_test): Add mi_version parameter, test -break-insert,
-break-info and =breakpoint-created.
gdb/doc/ChangeLog:
* gdb.texinfo (Mode Options): Mention mi3.
(Interpreters): Likewise.
(GDB/MI Development and Front Ends): Add entry for MI 3 in
version table. Document -fix-multi-location-breakpoint-output.
(GDB/MI Breakpoint Information): Document format of breakpoint
location output.
|
|
Changes in DejaGnu 1.6.2 mean that our testsuite will no longer run.
This is because of some confusion over how the gdb.exp file is
handled.
The gdb.exp file is really the tool init file, which is loaded from
within the DejaGnu core, and it should not be loaded directly from any
other file in the testsuite.
DejaGnu tries to prevent the same library being loaded twice by
remembering the names of library files as they are loaded. Until
recently loading the tool init file in DejaGnu was very similar to
loading a library file, as a result, loading the gdb.exp tool init
file simply recorded 'gdb.exp' as having been loaded, future attempts
to load 'gdb.exp' as a library would then be ignored (as the file was
marked as already loaded).
DejaGnu has now changed so that it supports having both a tool init
file and a library with the same name, something that was not possible
before. What this means however is that when the core loads the
'gdb.exp' tool init file it no longer marks the library 'gdb.exp' as
having been loaded. When we then execute 'load_lib gdb.exp' we then
try to reload the 'gdb.exp' file.
Unfortunately our gdb.exp file can only be loaded once. It use of
'rename cd builtin_cd' means that a second attempt to load this file
will fail.
This was discussed on the DejaGnu list here:
http://lists.gnu.org/archive/html/dejagnu/2019-03/msg00000.html
and the suggested advice is that, unless we have some real requirement
to load the tool init file twice, we should remove calls to 'load_lib
gdb.exp' and rely on DejaGnu to load the file for us, which is what
this patch does.
I've tested with native X86-64/GNU Linux and see no regressions.
gdb/testsuite/ChangeLog:
* config/default.exp: Remove 'load_lib gdb.exp'.
* config/monitor.exp: Likewise.
* config/sid.exp: Likewise.
* config/sim.exp: Likewise.
* config/slite.exp: Likewise.
* config/unix.exp: Likewise.
* gdb.base/default.exp: Remove unhelpful comment.
|
|
This patch adds support for the older TYPE*SIZE typenames that are
still around in older code.
For implementation this currently reuses the kind mechanism, as under
gFortran the kind number is equivalent to the size, however, this is
not necessarily true for all compilers. If the rules for other
compilers are better understood then this code might need to be
improved slightly to allow for a distinction between size and kind,
however, adding this extra complexity now seems pointless.
gdb/ChangeLog:
* f-exp.y (direct_abs_decl): Handle TYPE*SIZE type names.
gdb/testsuite/ChangeLog:
* gdb.fortran/type-kinds.exp: Extend to cover TYPE*SIZE cases.
|
|
Adds support for the abs intrinsic function, this requires adding a
new pattern to the Fortran parser. Currently only float and integer
argument types are supported to ABS, complex is still not supported,
this can be added later if needed.
gdb/ChangeLog:
* f-exp.y: New token, UNOP_INTRINSIC.
(exp): New pattern using UNOP_INTRINSIC token.
(f77_keywords): Add 'abs' keyword.
* f-lang.c: Add 'target-float.h' and 'math.h' includes.
(value_from_host_double): New function.
(evaluate_subexp_f): Support UNOP_ABS.
gdb/testsuite/ChangeLog:
* gdb.fortran/intrinsics.exp: Extend to cover ABS.
|
|
Switch to using TYPE_CODE_CHAR for character types. This appears to
have little impact on the test results as gFortran uses the
DW_TAG_string_type to represent all character variables (as far as I
can see). The only place this has an impact is when the user casts a
variable to a character type, in which case GDB does now use the CHAR
type, and prints the variable as both a value and a character, for
example, before:
(gdb) p ((character) 97)
$1 = 97
and after:
(gdb) p ((character) 97)
$1 = 97 'a'
gdb/ChangeLog:
* f-lang.c (build_fortran_types): Use TYPE_CODE_CHAR for character
types.
gdb/testsuite/ChangeLog:
* gdb.fortran/type-kinds.exp: Update expected results.
|
|
Add a new builtin type, an 8-byte integer, and allow GDB to parse
'integer (kind=8)', returning the new 8-byte integer.
gdb/ChangeLog:
* f-exp.y (convert_to_kind_type): Handle integer (kind=8).
* f-lang.c (build_fortran_types): Setup builtin_integer_s8.
* f-lang.h (struct builtin_f_type): Add builtin_integer_s8 field.
gdb/testsuite/ChangeLog:
* gdb.fortran/type-kinds.exp: Test new integer type kind.
|
|
Expand the number of types that can be adjusted with a (kind=N) type
extension.
gdb/ChangeLog:
* f-exp.y (convert_to_kind_type): Handle more type kinds.
gdb/testsuite/ChangeLog:
* gdb.fortran/type-kinds.exp (test_cast_1_to_type_kind): New
function.
(test_basic_parsing_of_type_kinds): Expand types tested.
(test_parsing_invalid_type_kinds): New function.
|
