This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH, LTO] add externally_visible attribute when necessary with -fwhole-program and resolution file.

> Yes, the above was a bug that was actually fixed. ?The issue that
> it chooses a random fortran common block and not the largest
> one still prevails.
>
> We might consider reverting the change on the trunk (it's certainly
> safe on the branch and makes more things work there). ?But it
> would also be nice for either gold or the linker-plugin being fixed
> for the size issue. ?(I think you run into the issue when building
> SPEC 2k6, you might also find a closed bugzilla that reports it)

Yes, gold merges common symbols by keeping the first and simply
increasing its size to the maximum of all subsequent ones. By the time
the plugin asks for symbol resolution information, there's no record
of which one was actually the largest. I'll work on a fix for this.

Is this in fact what PR 44149 is about?

-cary
Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]