[llvm-commits] [PATCH]Gold plugin: don't report error on ELF file

[Ivan Krasin]

FYI: I'm under impression that this patch fits to this point of the
LLVM Developer policy:

"""You are allowed to commit patches without approval which you think
are obvious. This is clearly a subjective decision — we simply expect
you to use good judgement. Examples include: fixing build breakage,
reverting obviously broken patches, documentation/comment changes, any
other minor changes."""

This patch is close to obvious and it fixes the build 'breakage' for
some systems (which are experienced a huge slowdown due to
hundreds/thousands of false warnings).

So, I would probably submit this patch tonight, if nobody complains about that.

krasin

On Mon, Sep 12, 2011 at 1:27 PM, Ivan Krasin <krasin at chromium.org> wrote:
> Hi llvm team,
>
> the patch is the follow up to r139340 (report errors occured in
> lto_module_create_from_*
> ), see the code review at
> http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20110905/127615.html
>
> That revision has turned on the diagnostics of failures inside LTO
> (before that, such failures were silently ignored by gold plugin). It
> appears that gold plugin did that for the reason: there's one
> "intended" fail: when the passed file is not a bitcode file, LTO fails
> to parse it and sets the error message appropriately. This does not
> look right. Gold plugin ought just look for the magic and skip
> non-bitcode files.
>
> This patch adds such a check. Also, it handles the case when the gold
> plugin is used not by gold, but by other tools like ar or nm, which
> don't currently provide get_view method.
>
> Please, take a look. Note: w/o this patch, r139340 has introduced
> several issues (like excessive false warnings) and if this patch looks
> completely wrong, it's probably better to revert r139340.
>
> The patch is attached to the email and is also available online:
> http://codereview.chromium.org/7844010/
>
> krasin
>