Created attachment 28649 [details] non-virtual-thunk.s This test previously passed, now it fails. A patch in the revision range (last_known_working:first_known_failing) 193333:193342 exposed or caused this regression. Since then it fails as follows: Running /tmp/hpautotest-gcc0/gcc/gcc/testsuite/g++.dg/debug/dwarf2/dwarf2.exp ... FAIL: g++.dg/debug/dwarf2/non-virtual-thunk.C -std=gnu++98 scan-assembler thunk.C:30 FAIL: g++.dg/debug/dwarf2/non-virtual-thunk.C -std=gnu++11 scan-assembler thunk.C:30 No extra clue in g++.log. Executing on host: /tmp/hpautotest-gcc0/cris-elf/gccobj/gcc/testsuite/g++/../../g++ -B/tmp/hpautotest-gcc0/cris-elf/gccobj/gcc/testsuite/g++/../../ /tmp/hpautotest-gcc0/gcc/gcc/testsuite/g++.dg/debug/dwarf2/non-virtual-thunk.C -fno-diagnostics-show-caret -nostdinc++ -I/tmp/hpautotest-gcc0/cris-elf/gccobj/cris-elf/libstdc++-v3/include/cris-elf -I/tmp/hpautotest-gcc0/cris-elf/gccobj/cris-elf/libstdc++-v3/include -I/tmp/hpautotest-gcc0/gcc/libstdc++-v3/libsupc++ -I/tmp/hpautotest-gcc0/gcc/libstdc++-v3/include/backward -I/tmp/hpautotest-gcc0/gcc/libstdc++-v3/testsuite/util -fmessage-length=0 -std=gnu++98 -g2 -dA -ffat-lto-objects -S -isystem /tmp/hpautotest-gcc0/cris-elf/gccobj/cris-elf/./newlib/targ-include -isystem /tmp/hpautotest-gcc0/gcc/newlib/libc/include -o non-virtual-thunk.s (timeout = 300) PASS: g++.dg/debug/dwarf2/non-virtual-thunk.C -std=gnu++98 (test for excess errors) FAIL: g++.dg/debug/dwarf2/non-virtual-thunk.C -std=gnu++98 scan-assembler thunk.C:30 (same for -std=gnu++11) There's no "thunk.C:30" in the assembly file, just lots of: "<TAB>;# /tmp/hpautotest-gcc0/gcc/gcc/testsuite/g++.dg/debug/dwarf2/non-virtual-thunk.C:33" See the attached non-virtual-thunk.s. Author of suspect patch in revision range CC:ed.
If the target doesn't call final_start_function and final_end_function in output_mi_thunk targhook (or equivalent), it is a target bug.
(In reply to comment #1) > If the target doesn't call final_start_function and final_end_function in > output_mi_thunk targhook (or equivalent), it is a target bug. I see. I either missed this requirement when I implemented those bits for CRIS or it wasn't a requirement at the time. BTW, I see several ports are missing those calls. I guess in the general case it can't be left to middle-end to call those functions at the right time, so I'll just update the port and hopefully the documentation, but no promises. Assigning the PR to myself while verifying that this is the cause.
It is a requirement if you want proper debug info or unwind info for the thunk, without that (or without calling the corresponding functions yourself, which is harder to maintain) you don't get proper debug info. It can't be called from the middle end, because some targets need to perform various things before final_start_function (or perhaps after final_end_function).
(In reply to comment #3) > It can't be called from > the middle end, because some targets need to perform various things before > final_start_function (or perhaps after final_end_function). Yes, that's what I meant, so let's just document the requirement.
Patch verified, changing component to target prior to commit.
Author: hp Date: Tue Nov 13 06:15:53 2012 New Revision: 193467 URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=193467 Log: PR target/55257 * config/cris/cris.c (cris_asm_output_mi_thunk): Call final_start_function and final_end_function. Modified: trunk/gcc/ChangeLog trunk/gcc/config/cris/cris.c
.