new abi file for x86_64-linux-gnu
Andreas Jaeger
aj@suse.de
Mon Jun 23 17:20:00 GMT 2003
Benjamin Kosnik <bkoz@redhat.com> writes:
>>I think everything has settled down - how shall we continue now?
>
> There are two things.
>
> 1) Right now, if there is no abi file, the output looks like:
>
> Baseline file doesn't exist.
> Try 'make new-abi-baseline' to create it.
> make[4]: [baseline_symbols] Error 1 (ignored)
> make[4]: [check-abi] Error 1 (ignored)
>
> Which seems to be confusing people. See:
>
> http://gcc.gnu.org/ml/gcc/2003-06/msg01837.html
There seems to be also another error: abi_check does not have
LD_LIBRARY_PATH setup correctly and therefore might access the wrong
library.
> it would be nice if this didn't error out, but instead just printed out
> the message about creating a new baseline file, and then silently
> skipped the check-abi rule. I don't have time to do this, but if
> somebody else does I'd appreciate it.
> 2) Updating the baseline files to 3.3.0 versions.
>
> First, the error where the "extract_symvers" script creates the current
> baseline symbols file without sorting it needs to be fixed. As it
> stands, the created file looks like it's got all kinds of difference
> from the 3.2.0 baseline, just because it's not sorted. This needs to be
> fixed.
>
> The updates should proceed as follows. If there is a 3.2.0 baseline,
> then get the gcc-3.3.0 sources, compile them, and do check-abi. If there
> are no incompatible symbols in the results, you can use the new baseline
> file as the update. (Note that the new-abi-baseline rule does the same
> thing.)
>
> If there is no 3.2.0 baseline file for a new target, it would be nice if
> it could be generated, then compared against the 3.3.0 baseline.
> However, it looks like this may be difficult or impossible for some
> targets (powerpc64?). So, I don't know what to do in this case. I guess
> just use the 3.3.0 generated file.
>
> Thoughts?
Looks reasonable. I'll try to look at this closer next weekend but if
anybody beats me to it, I wouldn't mind ;-)
Andreas
--
Andreas Jaeger, SuSE Linux AG, aj@suse.de, http://www.suse.de/~aj
GPG fingerprint = 93A3 365E CE47 B889 DF7F FED1 389A 563C C272 A126
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20030623/c660f9a7/attachment.sig>
More information about the Gcc-patches
mailing list