aboutsummaryrefslogtreecommitdiff
path: root/gdb/annotate.c
diff options
context:
space:
mode:
authorAndrew Burgess <andrew.burgess@embecosm.com>2020-05-12 21:29:23 +0100
committerAndrew Burgess <andrew.burgess@embecosm.com>2020-05-22 13:39:50 +0100
commit7c13f4e875fd5eeb0b7a1c301b4b513051822648 (patch)
treee1cd0f0c3fcb6892cea4a16c3bf3fe3823e51669 /gdb/annotate.c
parent6f3fe02b0b8f6865820d8889a8420190063f3235 (diff)
downloadgdb-7c13f4e875fd5eeb0b7a1c301b4b513051822648.zip
gdb-7c13f4e875fd5eeb0b7a1c301b4b513051822648.tar.gz
gdb-7c13f4e875fd5eeb0b7a1c301b4b513051822648.tar.bz2
gdb: Restore old annotations behaviour when printing frame info
This undoes most of the changes from these commits: commit ec8e2b6d3051f0b4b2a8eee9917898e95046c62f Date: Fri Jun 14 23:43:00 2019 +0100 gdb: Don't allow annotations to influence what else GDB prints commit 0d3abd8cc936360f8c46502135edd2e646473438 Date: Wed Jun 12 22:34:26 2019 +0100 gdb: Remove an update of current_source_line and current_source_symtab as a result of the discussion here: https://sourceware.org/pipermail/gdb/2020-April/048468.html Having taken time to reflect on the discussion, and reading the documentation again I believe we should revert GDB's behaviour back to how it used to be. The original concern that triggered the initial patch was that when annotations were on the current source and line were updated (inside the annotation code), while when annotations are off this update would not occur. This was incorrect, as printing the source with the call to print_source_lines does also update the current source and line. Further, the documentation here: https://sourceware.org/gdb/current/onlinedocs/gdb/Source-Annotations.html#Source-Annotations Clearly states: "The following annotation is used instead of displaying source code: ^Z^Zsource filename:line:character:middle:addr ..." So it is documented that the 'source' annotation is a replacement for, and not in addition to, actually printing the source lie. There are still a few issues that I can see, these are: 1. In source.c:info_line_command, when annotations are on we call annotate_source_line, however, if annotations are off then there is no corresponding call to print the source line. This means that a if a user uses 'info line ...' with annotations on, and then does a 'list', they will get different results than if they had done this with annotations off. 2. It bothers me that the call to annotate_source_line returns a boolean, and that this controls a call to print_source_line (in stack.c:print_frame_info). The reason for this is that the source line annotation will only print something if the file is found, and the line number is in range for the file. It seems to me like an annotation should always be printed, either one that identifies the file and line, or one that identifies the file and line GDB would like to access, but couldn't. I considered changing this, but in the end decided not too, if I extend the existing 'source' annotation to print something in all cases then I risk breaking existing UIs that rely on the file and line always being valid. If I add a new annotation then this might also break existing UIs that rely on GDB itself printing the error from within print_source_line. Given that annotations is deprecated (as I understand it) mechanism for UIs to interact with GDB (in favour of MI) I figure we should just restore the old behaviour, and leave the mini-bugs in until someone actually complains. This isn't a straight revert of the two commits mentioned above. I've left annotate_source_line instead of going back to the original identify_source_line, which lived in source.c, but was really annotation related. The API for setting the current source and line has changed since the original patches, so I updated for that change too. Finally I wrote the code in stack.c so that we avoided an extra level of indentation, which I felt made things easier to read. gdb/ChangeLog: * annotate.c (annotate_source_line): Update return type, add call to update current symtab and line. * annotate.h (annotate_source_line): Update return type, and extend header comment. * source.c (info_line_command): Check annotation_level before calling annotate_source_line. * stack.c (print_frame_info): If calling annotate_source_line returns true, then don't print any other source line information. gdb/testsuite/ChangeLog: * gdb.base/annota1.exp: Update expected results. * gdb.cp/annota2.exp: Update expected results, remove duplicate test name. * gdb.cp/annota3.exp: Update expected results.
Diffstat (limited to 'gdb/annotate.c')
-rw-r--r--gdb/annotate.c19
1 files changed, 14 insertions, 5 deletions
diff --git a/gdb/annotate.c b/gdb/annotate.c
index 6daa0c5..0a4e2f2 100644
--- a/gdb/annotate.c
+++ b/gdb/annotate.c
@@ -435,7 +435,7 @@ annotate_source (const char *filename, int line, int character, int mid,
/* See annotate.h. */
-void
+bool
annotate_source_line (struct symtab *s, int line, int mid_statement,
CORE_ADDR pc)
{
@@ -443,16 +443,25 @@ annotate_source_line (struct symtab *s, int line, int mid_statement,
{
const std::vector<off_t> *offsets;
if (!g_source_cache.get_line_charpos (s, &offsets))
- return;
-
- /* Don't index off the end of the line_charpos array. */
+ return false;
if (line > offsets->size ())
- return;
+ return false;
annotate_source (s->fullname, line, (int) (*offsets)[line - 1],
mid_statement, SYMTAB_OBJFILE (s)->arch (),
pc);
+
+ /* Update the current symtab and line. */
+ symtab_and_line sal;
+ sal.pspace = SYMTAB_PSPACE (s);
+ sal.symtab = s;
+ sal.line = line;
+ set_current_source_symtab_and_line (sal);
+
+ return true;
}
+
+ return false;
}