Control: reassign -1 gdb 9.1-1 Control: forwarded -1 https://sourceware.org/ml/gdb-patches/2020-03/msg00010.html Control: retitle -1 [regression] GDB can't printf convenience vars that hold pointers to strings
On Wednesday, February 19 2020, gregor herrmann wrote: > On Wed, 19 Feb 2020 14:16:52 +0100, Héctor Orón Martínez wrote: > >> This issue is now affecting GDB 9.1 autopkgtests and preventing it from >> migrating to testing. >> >> https://ci.debian.net/data/autopkgtest/testing/amd64/liba/libapp-stacktrace-perl/4328197/log.gz >> >> https://ci.debian.net/data/autopkgtest/testing/arm64/liba/libapp-stacktrace-perl/4314744/log.gz > > Oh, one of the bugs which "hide": > https://ci.debian.net/packages/liba/libapp-stacktrace-perl/testing/amd64/ > is all green, and after clicking on britney it turns red: > https://ci.debian.net/user/britney/jobs?package=libapp-stacktrace-perl&suite[]=testing&arch[]=amd64 Hi there, After spending a non-trivial amount of time investigating this over the weekend, I was able to find the underlying case of the bug (and come up with a fix). Please, keep reading. > Anyway, in the meantime it also FTBFS in unstable: > > # [Thread debugging using libthread_db enabled] > # Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1". > # 0x00007f7b77cc1047 in select () from /lib/x86_64-linux-gnu/libc.so.6 > # à|ÚU:101 > # à|ÚU:106 > # à|ÚU:106 > # à|ÚU:106 > # à|ÚU:106 > # à|ÚU:106 > # à|ÚU:106 > # à|ÚU:106 > # à|ÚU:106 > # à|ÚU:106 > # à|ÚU:106 > # [Inferior 1 (process 25884) detached] > > # Failed test at t/unthreaded.t line 55. > # '[Thread debugging using libthread_db enabled] > # Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1". > # 0x00007f7b77cc1047 in select () from /lib/x86_64-linux-gnu/libc.so.6 > # à|ÚU:101 > # à|ÚU:106 > # à|ÚU:106 > # à|ÚU:106 > # à|ÚU:106 > # à|ÚU:106 > # à|ÚU:106 > # à|ÚU:106 > # à|ÚU:106 > # à|ÚU:106 > # à|ÚU:106 > # [Inferior 1 (process 25884) detached] > # ' > # doesn't match '(?^mx: > # (?: > # ^t/unthreaded\.t:\d+\n > # ){10} > # )' > # Looks like you failed 1 test of 5. > t/unthreaded.t ............ > 1..5 > not ok 1 > ok 2 - Reaped pstack > ok 3 - exit(0) > ok 4 - No signals > ok 5 - No core dump > # Alive top 1582127930: 1 middle: 1 > # Alive top 1582127931: 1 middle: 0 > Dubious, test returned 1 (wstat 256, 0x100) > Failed 1/5 subtests > > Test Summary Report > ------------------- > t/unthreaded.t (Wstat: 256 Tests: 5 Failed: 1) > Failed test: 1 > Non-zero exit status: 1 > Files=4, Tests=6, 2 wallclock secs ( 0.03 usr 0.01 sys + 0.96 cusr 0.09 > csys = 1.09 CPU) > Result: FAIL > > (I _think_ that's yet another different error.) This seems to be the same failure reported by doko. > In testing with gdb 8.3.1 the package still builds: > > # [Thread debugging using libthread_db enabled] > # Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1". > # 0x00007f1b9eb77047 in select () from /lib/x86_64-linux-gnu/libc.so.6 > # t/unthreaded.t:101 > # t/unthreaded.t:106 > # t/unthreaded.t:106 > # t/unthreaded.t:106 > # t/unthreaded.t:106 > # t/unthreaded.t:106 > # t/unthreaded.t:106 > # t/unthreaded.t:106 > # t/unthreaded.t:106 > # t/unthreaded.t:106 > # t/unthreaded.t:106 > # [Inferior 1 (process 3207) detached] > t/unthreaded.t ............ > 1..5 > ok 1 > ok 2 - Reaped pstack > ok 3 - exit(0) > ok 4 - No signals > ok 5 - No core dump > # Alive top 1582128303: 1 middle: 1 > # Alive top 1582128304: 1 middle: 0 > ok Right. This is actually a regression that happened on GDB (therefore I'm reassigning the bug). So, it took me quite a bit of time to debug this because libapp-stacktrace-perl doesn't help when it comes to seeing what GDB is doing inside it... But I was able to determine that the problem started happening after the following upstream commit: commit 1f6f6e21fa86dc3411a6498608f32e9eb24b7851 Author: Philippe Waroquiers <[email protected]> Date: Mon Jun 10 21:41:51 2019 +0200 Ensure GDB printf command can print convenience var strings without a target. As the commit title says, the goal here was to make GDB's "printf" command be able to print convenience variables that are strings without the need for a target (a program/corefile being debugged, for example). However, the commit changed the way GDB treats convenience variables that *point* to addresses referring to e.g. a "char *" inside the debuggee's memory space. GDB will treat this as a string (because it's pointing to a "char *"), and will try to obtain its value via the wrong mechanism. I won't dive too deep here; suffice it to say that if we treat convenience variables that hold pointers in a different manner, then things get back to normal. I submitted a tentative patch to fix this upstream, and will backport it downstream once my approach is deemed correct. Thanks, -- Sergio GPG key ID: 237A 54B1 0287 28BF 00EF 31F4 D0EB 7628 65FC 5E36 Please send encrypted e-mail if possible http://sergiodj.net/
signature.asc
Description: PGP signature
