firefox 68.0{,.1} i686 needs a patch to compile

[tnut]

Following patch ​https://bugs.chromium.org/p/skia/issues/detail?id=9202 should be apply when compiling on 32 bits machines

[Douglas R. Reno]

Hi tnut, did you manage to bypass the rust error that causes compilation to fail due to not being able to allocate enough RAM? Even with swap, it seems to refuse to use it.

Either way though, I can add a patch for this.

[Douglas R. Reno]

I think a sed might be more appropriate though, with a case statement wrapping it:

case $(uname -m) in

i?86) sed -i ...

esac

[Douglas R. Reno]

Triaging: move to Defect.

[tnut]

to avoid the memory problem, you should add in your mozconfig:

ac_add_options --disable-debug-symbols

This Apply to Thunderbird as well.

I aggree for the sed, nevertheless here is the patch I did use:

​https://github.com/NuTyX/packages-i686/blob/rolling/gui-extra/firefox/firefox-68.0.1-skcms-1.patch

[Douglas R. Reno]

Now that 68.0.2 is out, I'll test it with that version before committing.

Here's what I have so far:

Add a commented "ac_add_options --disable-debug-symbols" to mozconfig, otherwise gkrust will fail to link (llvm complains about being out of memory). Description for that will be something like "If you are compiling on i686, gkrust will fail to link due to memory limits. To fix this, disable debug symbols by uncommenting the line below."

Add a sed command to fix an internal compiler error caused by inlining functions on GCC-7 and later:

sed -i "562 s/mips64/i386/" gfx/skia/skia/third_party/skcms/src/Transform_inl.h

These two allowed a build of 68.0.1 to pass for me last night.

[Douglas R. Reno]

Fixed at r21978