On Wed, Jun 1, 2022 at 6:32 AM Richard Purdie <richard.pur...@linuxfoundation.org> wrote: > > On Wed, 2022-06-01 at 06:21 -1000, Steve Sakoman wrote: > > On Wed, Jun 1, 2022 at 6:10 AM Richard Purdie > > <richard.pur...@linuxfoundation.org> wrote: > > > > > > On Wed, 2022-06-01 at 05:29 -1000, Steve Sakoman wrote: > > > > > > > > > Keep in mind that the test uses sstate to compare against so even that > > > isn't a guarantee of identifying the issue. > > > > > > Did you try the strings comparison I suggested? Do you have the file > > > section tables comparison (diff) I could look at? > > > > Are you referring to the section header info from objdump -h ? > > > > If so I've attached the a/b info, they are quite compact. > > Yes, I couldn't remember the option to objdump offhand. The diff looks > like: > > --- /tmp/headersa.txt 2022-06-01 17:22:57.272277627 +0100 > +++ /tmp/headersb.txt 2022-06-01 17:22:43.972218408 +0100 > @@ -1,5 +1,5 @@ > > -a/usr/lib/libwebkit2gtk-4.0.so.37.56.5: file format elf64-x86-64 > +b/usr/lib/libwebkit2gtk-4.0.so.37.56.5: file format elf64-x86-64 > > Sections: > Idx Name Size VMA LMA File off > Algn > @@ -63,15 +63,15 @@ > CONTENTS, READONLY, DEBUGGING, OCTETS > 29 .debug_abbrev 00c6ce64 0000000000000000 0000000000000000 921c9968 > 2**0 > CONTENTS, READONLY, DEBUGGING, OCTETS > - 30 .debug_line 09025ee1 0000000000000000 0000000000000000 92e367cc > 2**0 > + 30 .debug_line 09025ede 0000000000000000 0000000000000000 92e367cc > 2**0 > CONTENTS, READONLY, DEBUGGING, OCTETS > - 31 .debug_str 1820ba50 0000000000000000 0000000000000000 9be5c6ad > 2**0 > + 31 .debug_str 1820ba50 0000000000000000 0000000000000000 9be5c6aa > 2**0 > CONTENTS, READONLY, DEBUGGING, OCTETS > - 32 .debug_line_str 0005d9f4 0000000000000000 0000000000000000 b40680fd > 2**0 > + 32 .debug_line_str 0005d9f4 0000000000000000 0000000000000000 b40680fa > 2**0 > CONTENTS, READONLY, DEBUGGING, OCTETS > - 33 .debug_loclists 13d74aa2 0000000000000000 0000000000000000 b40c5af1 > 2**0 > + 33 .debug_loclists 13d74aa0 0000000000000000 0000000000000000 b40c5aee > 2**0 > CONTENTS, READONLY, DEBUGGING, OCTETS > - 34 .debug_rnglists 034c6c08 0000000000000000 0000000000000000 c7e3a593 > 2**0 > + 34 .debug_rnglists 034c6bf6 0000000000000000 0000000000000000 c7e3a58e > 2**0 > CONTENTS, READONLY, DEBUGGING, OCTETS > - 35 .gnu_debuglink 00000024 0000000000000000 0000000000000000 cb30119c > 2**2 > + 35 .gnu_debuglink 00000024 0000000000000000 0000000000000000 cb301184 > 2**2 > CONTENTS, READONLY > > I'm a bit puzzled since usually the debug symbols are moved off to the > -dbg package and yet they seem combined here with the library. What the > above seems to say is that the debug_line, loclists and rnglists > sections changed size, the rest changed offset (and we ignore > gnu_debuglink since that is a checksum which would depend on the other > bits). > > That would imply that some piece of code used to compile webkit has > different line numbering on one system compared to another but > generates identical code. You should be able to parse the debuginfo to > get a filename and linenumber out of it but I can't remember how, I > think I learnt to read dwarfish last time I did this :/.
I did a dump of the debug sections with objdump -g and got absolutely huge output files that crashed all of the diff programs I tried. I guess the next step is to split those files into smaller chunks and diff those :-( Have I mentioned how much I hate webkit-gtk? Steve
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#166437): https://lists.openembedded.org/g/openembedded-core/message/166437 Mute This Topic: https://lists.openembedded.org/mt/91339755/21656 Group Owner: openembedded-core+ow...@lists.openembedded.org Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-
