# This testcase is part of GDB, the GNU debugger. # Copyright 2004, 2007, 2008, 2009, 2010 Free Software Foundation, Inc. # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 3 of the License, or # (at your option) any later version. # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # You should have received a copy of the GNU General Public License # along with this program. If not, see . # Test "return", "finish", and "call" of functions that a scalar (int, # float, enum) and/or take a single scalar parameter. if $tracelevel then { strace $tracelevel } # Some targets can't call functions, so don't even bother with this # test. if [target_info exists gdb,cannot_call_functions] { setup_xfail "*-*-*" fail "This target can not call functions" continue } set testfile "call-sc" set srcfile ${testfile}.c set binfile ${objdir}/${subdir}/${testfile} # Create and source the file that provides information about the # compiler used to compile the test case. if [get_compiler_info ${binfile}] { return -1; } # Compile a variant of scalars.c using TYPE to specify the type of the # parameter and return-type. Run the compiled program up to "main". # Also updates the global "testfile" to reflect the most recent build. proc start_scalars_test { type } { global testfile global srcfile global binfile global objdir global subdir global srcdir global gdb_prompt global expect_out # Create the additional flags set flags "debug additional_flags=-DT=${type}" set testfile "call-sc-${type}" set binfile ${objdir}/${subdir}/${testfile} if { [gdb_compile "${srcdir}/${subdir}/${srcfile}" "${binfile}" executable "${flags}"] != "" } { # built the second test case since we can't use prototypes warning "Prototypes not supported, rebuilding with -DNO_PROTOTYPES" if { [gdb_compile "${srcdir}/${subdir}/${srcfile}" "${binfile}" executable "${flags} additional_flags=-DNO_PROTOTYPES"] != "" } { untested call-sc.exp return -1 } } # Start with a fresh gdb. gdb_exit gdb_start gdb_reinitialize_dir $srcdir/$subdir gdb_load ${binfile} # Make certain that the output is consistent gdb_test "set print sevenbit-strings" "" \ "set print sevenbit-strings; ${testfile}" gdb_test "set print address off" "" \ "set print address off; ${testfile}" gdb_test "set width 0" "" \ "set width 0; ${testfile}" # Advance to main if { ![runto_main] } then { gdb_suppress_tests; } # Get the debug format get_debug_format # check that type matches what was passed in set test "ptype; ${testfile}" set foo_t "xxx" gdb_test_multiple "ptype ${type}" "${test}" { -re "type = (\[^\r\n\]*)\r\n$gdb_prompt $" { set foo_t "$expect_out(1,string)" pass "$test (${foo_t})" } } gdb_test "ptype foo" "type = ${foo_t}" "ptype foo; ${testfile} $expect_out(1,string)" } # Given N (0..25), return the corresponding alphabetic letter in lower # or upper case. This is ment to be i18n proof. proc i2a { n } { return [string range "abcdefghijklmnopqrstuvwxyz" $n $n] } proc I2A { n } { return [string toupper [i2a $n]] } # Test GDB's ability to make inferior function calls to functions # returning (or passing) in a single scalar. # start_scalars_test() will have previously built a program with a # specified scalar type. To ensure robustness of the output, "p/c" is # used. # This tests the code paths "which return-value convention?" and # "extract return-value from registers" called by "infcall.c". proc test_scalar_calls { } { global testfile global gdb_prompt # Check that GDB can always extract a scalar-return value from an # inferior function call. Since GDB always knows the location of # an inferior function call's return value these should never fail # Implemented by calling the parameterless function "fun" and then # examining the return value printed by GDB. set tests "call ${testfile}" # Call fun, checking the printed return-value. gdb_test "p/c fun()" "= 49 '1'" "p/c fun(); ${tests}" # Check that GDB can always pass a structure to an inferior function. # This test can never fail. # Implemented by calling the one parameter function "Fun" which # stores its parameter in the global variable "L". GDB then # examining that global to confirm that the value is as expected. gdb_test "call Fun(foo)" "" "call Fun(foo); ${tests}" gdb_test "p/c L" " = 49 '1'" "p/c L; ${tests}" } # Test GDB's ability to both return a function (with "return" or # "finish") and correctly extract/store any corresponding # return-value. # Check that GDB can consistently extract/store structure return # values. There are two cases - returned in registers and returned in # memory. For the latter case, the return value can't be found and a # failure is "expected". However GDB must still both return the # function and display the final source and line information. # N identifies the number of elements in the struct that will be used # for the test case. FAILS is a list of target tuples that will fail # this test. # This tests the code paths "which return-value convention?", "extract # return-value from registers", and "store return-value in registers". # Unlike "test struct calls", this test is expected to "fail" when the # return-value is in memory (GDB can't find the location). The test # is in three parts: test "return"; test "finish"; check that the two # are consistent. GDB can sometimes work for one command and not the # other. proc test_scalar_returns { } { global gdb_prompt global testfile set tests "return ${testfile}" # Check that "return" works. # GDB must always force the return of a function that has # a struct result. Dependant on the ABI, it may, or may not be # possible to store the return value in a register. # The relevant code looks like "L{n} = fun{n}()". The test forces # "fun{n}" to "return" with an explicit value. Since that code # snippet will store the the returned value in "L{n}" the return # is tested by examining "L{n}". This assumes that the # compiler implemented this as fun{n}(&L{n}) and hence that when # the value isn't stored "L{n}" remains unchanged. Also check for # consistency between this and the "finish" case. # Get into a call of fun gdb_test "advance fun" \ "fun .*\[\r\n\]+\[0-9\].*return foo.*" \ "advance to fun for return; ${tests}" # Check that the program invalidated the relevant global. gdb_test "p/c L" " = 90 'Z'" "zed L for return; ${tests}" # Force the "return". This checks that the return is always # performed, and that GDB correctly reported this to the user. # GDB 6.0 and earlier, when the return-value's location wasn't # known, both failed to print a final "source and line" and misplaced # the frame ("No frame"). # The test is writen so that it only reports one FAIL/PASS for the # entire operation. The value returned is checked further down. # "return_value_unknown", if non-empty, records why GDB realised # that it didn't know where the return value was. set test "return foo; ${tests}" set return_value_unknown 0 set return_value_unimplemented 0 gdb_test_multiple "return foo" "${test}" { -re "The location" { # Ulgh, a struct return, remember this (still need prompt). set return_value_unknown 1 exp_continue } -re "A structure or union" { # Ulgh, a struct return, remember this (still need prompt). set return_value_unknown 1 # Double ulgh. Architecture doesn't use return_value and # hence hasn't implemented small structure return. set return_value_unimplemented 1 exp_continue } -re "Make fun return now.*y or n. $" { gdb_test_multiple "y" "${test}" { -re "L *= fun.*${gdb_prompt} $" { # Need to step off the function call gdb_test "next" "zed.*" "${test}" } -re "zed \\(\\);.*$gdb_prompt $" { pass "${test}" } } } } # If the previous test did not work, the program counter might # still be inside foo() rather than main(). Make sure the program # counter is is main(). # # This happens on ppc64 GNU/Linux with gcc 3.4.1 and a buggy GDB set test "return foo; synchronize pc to main()" for {set loop_count 0} {$loop_count < 2} {incr loop_count} { gdb_test_multiple "backtrace 1" $test { -re "#0.*main \\(\\).*${gdb_prompt} $" { pass $test set loop_count 2 } -re "#0.*fun \\(\\).*${gdb_prompt} $" { if {$loop_count < 1} { gdb_test "finish" ".*" "" } else { fail $test set loop_count 2 } } } } # Check that the return-value is as expected. At this stage we're # just checking that GDB has returned a value consistent with # "return_value_unknown" set above. set test "value foo returned; ${tests}" gdb_test_multiple "p/c L" "${test}" { -re " = 49 '1'.*${gdb_prompt} $" { if $return_value_unknown { # This contradicts the above claim that GDB didn't # know the location of the return-value. fail "${test}" } else { pass "${test}" } } -re " = 90 .*${gdb_prompt} $" { if $return_value_unknown { # The struct return case. Since any modification # would be by reference, and that can't happen, the # value should be unmodified and hence Z is expected. # Is this a reasonable assumption? pass "${test}" } else { # This contradicts the above claim that GDB knew # the location of the return-value. fail "${test}" } } -re ".*${gdb_prompt} $" { if $return_value_unimplemented { # What a suprize. The architecture hasn't implemented # return_value, and hence has to fail. kfail "$test" gdb/1444 } else { fail "$test" } } } # Check that a "finish" works. # This is almost but not quite the same as "call struct funcs". # Architectures can have subtle differences in the two code paths. # The relevant code snippet is "L{n} = fun{n}()". The program is # advanced into a call to "fun{n}" and then that function is # finished. The returned value that GDB prints, reformatted using # "p/c", is checked. # Get into "fun()". gdb_test "advance fun" \ "fun .*\[\r\n\]+\[0-9\].*return foo.*" \ "advance to fun for finish; ${tests}" # Check that the program invalidated the relevant global. gdb_test "p/c L" " = 90 'Z'" "zed L for finish; ${tests}" # Finish the function, set 'finish_value_unknown" to non-empty if the # return-value was not found. set test "finish foo; ${tests}" set finish_value_unknown 0 gdb_test_multiple "finish" "${test}" { -re "Value returned is .*${gdb_prompt} $" { pass "${test}" } -re "Cannot determine contents.*${gdb_prompt} $" { # Expected bad value. For the moment this is ok. set finish_value_unknown 1 pass "${test}" } } # Re-print the last (return-value) using the more robust # "p/c". If no return value was found, the 'Z' from the previous # check that the variable was cleared, is printed. set test "value foo finished; ${tests}" gdb_test_multiple "p/c" "${test}" { -re " = 49 '1'\[\r\n\]+${gdb_prompt} $" { if $finish_value_unknown { # This contradicts the above claim that GDB didn't # know the location of the return-value. fail "${test}" } else { pass "${test}" } } -re " = 90 'Z'\[\r\n\]+${gdb_prompt} $" { # The value didn't get found. This is "expected". if $finish_value_unknown { pass "${test}" } else { # This contradicts the above claim that GDB did # know the location of the return-value. fail "${test}" } } } # Finally, check that "return" and finish" have consistent # behavior. # Since both "return" and "finish" use equivalent "which # return-value convention" logic, both commands should have # identical can/can-not find return-value messages. # Note that since "call" and "finish" use common code paths, a # failure here is a strong indicator of problems with "store # return-value" code paths. Suggest looking at "return_value" # when investigating a fix. set test "return and finish use same convention; ${tests}" if {$finish_value_unknown == $return_value_unknown} { pass "${test}" } else { kfail gdb/1444 "${test}" } } # ABIs pass anything >8 or >16 bytes in memory but below that things # randomly use register and/and structure conventions. Check all # possible sized char scalars in that range. But only a restricted # range of the other types. # NetBSD/PPC returns "unnatural" (3, 5, 6, 7) sized scalars in memory. # d10v is weird. 5/6 byte scalars go in memory. 2 or more char # scalars go in memory. Everything else is in a register! # Test every single char struct from 1..17 in size. This is what the # original "scalars" test was doing. start_scalars_test tc test_scalar_calls test_scalar_returns # Let the fun begin. # Assuming that any integer struct larger than 8 bytes goes in memory, # come up with many and varied combinations of a return struct. For # "struct calls" test just beyond that 8 byte boundary, for "struct # returns" test up to that boundary. # For floats, assumed that up to two struct elements can be stored in # floating point registers, regardless of their size. # The approx size of each structure it is computed assumed that tc=1, # ts=2, ti=4, tl=4, tll=8, tf=4, td=8, tld=16, and that all fields are # naturally aligned. Padding being added where needed. Note that # these numbers are just approx, the d10v has ti=2, a 64-bit has has # tl=8. # Approx size: 2, 4, ... start_scalars_test ts test_scalar_calls test_scalar_returns # Approx size: 4, 8, ... start_scalars_test ti test_scalar_calls test_scalar_returns # Approx size: 4, 8, ... start_scalars_test tl test_scalar_calls test_scalar_returns # Approx size: 8, 16, ... start_scalars_test tll test_scalar_calls test_scalar_returns # Approx size: 4, 8, ... start_scalars_test tf test_scalar_calls test_scalar_returns # Approx size: 8, 16, ... start_scalars_test td test_scalar_calls test_scalar_returns # Approx size: 16, 32, ... start_scalars_test tld test_scalar_calls test_scalar_returns # Approx size: 4, 8, ... start_scalars_test te test_scalar_calls test_scalar_returns return 0