Bug 146817 – patches for gcc-4.1.1 to fix visibility-related bugs

• Product: Gentoo Linux
• Status: RESOLVED UPSTREAM

[Description Jan Jitse Venselaar 2006-09-08 03:08:08 UTC]

I modified some patches applied to upstream gcc-4.2 branch and redhat gcc-4.1 branch which fix a lot of visibility related bugs.
I took the patches from the redhat gcc-4.1 branch, and removed the changelog parts (did not apply) and modified them to apply to Gentoo gcc-4.1.1-r1

They fix the following bugs:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17470
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19134
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20297
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21581
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21675
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25915
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26612
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26905
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26984
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27000
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28215
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28279

See http://gcc.gnu.org/ml/gcc-patches/2006-06/msg01302.html for a description of the patches. Upstream will most probably not apply these patches to gcc-4.1 branch, but maybe Gentoo can do this? I compiled most of KDE (everything I have installed) with these patches and kdehiddenvisibility, and nothing visibly broke. Also, without these patches I was not able to compile geki's openoffice ebuilds, with these patches I was able to. (see http://forums.gentoo.org/viewtopic-t-489909-start-50.html )

[Comment 1 Jan Jitse Venselaar 2006-09-08 03:08:44 UTC]

Created attachment 96372 [details, diff]
first patch

[Comment 2 Jan Jitse Venselaar 2006-09-08 03:09:13 UTC]

Created attachment 96373 [details, diff]
broken of from previous patch in order to apply

[Comment 3 Jan Jitse Venselaar 2006-09-08 03:09:47 UTC]

Created attachment 96374 [details, diff]
second patch

[Comment 4 Jan Jitse Venselaar 2006-09-08 03:10:08 UTC]

Created attachment 96375 [details, diff]
third patch

[Comment 5 Diego Elio Pettenò (RETIRED) 2006-09-12 12:25:03 UTC]

KDE does not require these patches (builds and work fine without them), but there are other cases where these patches can break, will try if these fixes the one I have locally.

[Comment 6 Diego Elio Pettenò (RETIRED) 2006-09-12 14:05:04 UTC]

This does not fix bug #134162.

[Comment 7 Diego Elio Pettenò (RETIRED) 2006-09-18 19:41:36 UTC]

Warning: this might be breaking more visibility support. I've started getting text relocations on FFmpeg where it used to work fine. I'll let you know for sure after a gcc rebuild.

[Comment 8 Diego Elio Pettenò (RETIRED) 2006-09-19 07:58:46 UTC]

Sorry, false warning; visibility support did not change a bit with or without these patches for the software I can test myself.

[Comment 9 Jan Jitse Venselaar 2006-09-25 11:28:43 UTC]

Hmm... I get a compile failure with these patches and mozilla-firefox and mozilla-thunderbird. Might not be worthy of inclusion after all, but maybe the error really is in firefox? Anyone got a clue?


nsXPCOMObsolete.o:(.data.rel.ro+0x20): undefined reference to `nsFileSpecImpl::Create(nsISupports*, nsID const&, void**)'
nsXPCOMObsolete.o:(.data.rel.ro+0x80): undefined reference to `nsDirectoryIteratorImpl::Create(nsISupports*, nsID const&, void**)'
/usr/lib/gcc/x86_64-pc-linux-gnu/4.1.1/../../../../x86_64-pc-linux-gnu/bin/ld: libxpcom_compat_c.so: hidden symbol `nsFileSpecImpl::Create(nsISupports*, nsID const&, void**)' isn't defined
/usr/lib/gcc/x86_64-pc-linux-gnu/4.1.1/../../../../x86_64-pc-linux-gnu/bin/ld: final link failed: Nonrepresentable section on output
collect2: ld returned 1 exit status
gmake[3]: *** [libxpcom_compat_c.so] Error 1
gmake[3]: Leaving directory `/var/tmp/portage/mozilla-firefox-1.5.0.7/work/mozilla/xpcom/obsolete/component'
gmake[2]: *** [libs] Error 2

[Comment 10 Ryan Hill (RETIRED) 2006-10-21 19:06:50 UTC]

not that i have any say in this ;), but i think it'd be better to wait for 4.2 than have to maintain these ourselves.  they're fairly invasive and visibility handling got a major overhaul for 4.2.  it seems like a bad idea to completely change how visibility works in the middle of a stable series.