This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: [PATCH] Use `supports_gnu_unique' with the `unique_symbol' and `type' tests
- From: Nick Clifton <nickc at redhat dot com>
- To: "Maciej W. Rozycki" <macro at imgtec dot com>
- Cc: binutils at sourceware dot org
- Date: Tue, 28 Jun 2016 13:51:08 +0100
- Subject: Re: [PATCH] Use `supports_gnu_unique' with the `unique_symbol' and `type' tests
- Authentication-results: sourceware.org; auth=none
- References: <alpine dot DEB dot 2 dot 00 dot 1606281151340 dot 4103 at tp dot orcam dot me dot uk>
Hi Maciej,
> This has been regression-tested against the usual 162 targets. OK to
> apply?
Yes, please do.
> While testing this I've noticed an `alpha-unknown-freebsd4.7' progression
> has gone and this is due to your recent change with commit 96037eb0b8c2
> ("Fix various binutils testsuite failures.") to exclude `alpha-*-*' from
> `ar.exp' testing. I think it is too broad, I think only non-ELF (ECOFF)
> Alpha targets might have problems with `ar.exp' (except for the GNU unique
> feature, handled here) as these tests used to pass for ELF targets like
> `alpha-linux', `alpha-netbsd' or said `alpha-unknown-freebsd4.7'. So I
> think you have unnecessarily reduced coverage (assuming that any failures
> for ECOFF were due to a shortcoming of that format rather than our bug).
>
> I think updating the condition to say:
>
> if { [istarget "alpha-*-*"] && ![is_elf_format] } then {
> return
> }
>
> will do, however, unfortunately, I'm still busy with 2.27 clean-ups, so I
> can't even afford machine cycles for testing this refined update to your
> sledgehammer approach -- will you therefore please look into it yourself?
Will do, and thanks for pointing this out.
Cheers
Nick