This is the mail archive of the mailing list for the binutils 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: Release 2.23: Ping

On Thu, 13 Sep 2012, Matthias Klose wrote:

> >  Thanks a lot indeed!  I'll just regression-test it against my latest 
> > trunk results and see if anything pops up.  I'll let you know the outcome 
> > by tomorrow.
> this did break linux kernel builds (at least in 3.5.0), on x86_64-linux-gnu.
> Reverting Tristan's checkin (git 754bb8efffca0d3006e417ba56b92bec9990753d) of
> the mips backports to the 2.23 branch solves the issue. The patch touches
> generic files in bfd and ld. I didn't check, if it's a binutils issue or a
> kernel issue exposed by this change.

 Some changes were indeed generic, specifically the symbol export class 
BFD bug fix.  The bug affected all targets, however Linux shouldn't care 
as export classes are used in dynamic linking only.

 There were also non-MIPS changes to ld that were supposed to be 
transparent; they mainly affected symbols defined by linker scripts and 
caused no test suite regressions -- but then that's a feature used quite 
heavily by Linux.  If anything was indeed missed, then it would be a 
worthwhile addition to the test suite alongside the fix.

> From
> """
> $ readelf -s setup.elf | grep video_card
>    138: 00003660 0 NOTYPE GLOBAL DEFAULT 12 video_cards
>    151: 00003660 0 NOTYPE GLOBAL DEFAULT 12 video_cards_end
> We should see some space between video_cards and video_cards_end, and the fact
> that we don't means the kernel isn't going to find any options for video.
> """

 Thanks for the reference.  Can you please give me some instructions as to 
how to build a kernel that has this problem, e.g. will some defconfig and 
a tree be enough to reproduce that?  I'll have a look into it, 
but please forgive my lack of knowledge on the recipes Ubuntu developers 
use.  Thanks.

 Also would you per any chance be willing to help and try binutils trunk 
to track down the specific change that caused this problem?  The backport 
was made in bulk while on trunk self-contained changes were applied 
individually.  That would make tracking down the culprit much much easier.


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