maintainer-feedback requested: [Bug 270782] www/firefox: 112 fails to build on i386
- In reply to: bugzilla-noreply_a_freebsd.org: "[Bug 270782] www/firefox: 112 fails to build on i386"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Wed, 12 Apr 2023 09:18:41 UTC
Bugzilla Automation <bugzilla@FreeBSD.org> has asked freebsd-gecko (Nobody)
<gecko@FreeBSD.org> for maintainer-feedback:
Bug 270782: www/firefox: 112 fails to build on i386
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=270782
--- Description ---
Building on i386 fails with these errors:
---
In file included from Unified_cpp_dom_base10.cpp:74:
In file included from
/wrkdirs/usr/ports/www/firefox/work/firefox-112.0/dom/base/nsTextFragment.cpp:1
62:
/wrkdirs/usr/ports/www/firefox/work/firefox-112.0/dom/base/nsTextFragmentGeneri
c.h:16:48:
error: implicit instantiation of undefined
template 'xsimd::batch<short, xsimd::sse2>'
const uint32_t numUnicharsPerVector = xsimd::batch<int16_t, Arch>::size;
^
/wrkdirs/usr/ports/www/firefox/work/firefox-112.0/dom/base/nsTextFragment.cpp:1
83:21:
note: in instantiation of function template sp
ecialization 'mozilla::FirstNon8Bit<xsimd::sse2>' requested here
return mozilla::FirstNon8Bit<xsimd::sse2>(str, end);
^
/wrkdirs/usr/ports/www/firefox/work/firefox-112.0/third_party/xsimd/include/xsi
md/arch/xsimd_scalar.hpp:30:11:
note: template is dec
lared here
class batch;
^
In file included from Unified_cpp_dom_base10.cpp:74:
In file included from
/wrkdirs/usr/ports/www/firefox/work/firefox-112.0/dom/base/nsTextFragment.cpp:1
62:
/wrkdirs/usr/ports/www/firefox/work/firefox-112.0/dom/base/nsTextFragmentGeneri
c.h:35:31:
error: implicit instantiation of undefined
template 'xsimd::batch<short, xsimd::sse2>'
xsimd::batch<int16_t, Arch> vectmask(static_cast<int16_t>(shortMask));
^
/wrkdirs/usr/ports/www/firefox/work/firefox-112.0/third_party/xsimd/include/xsi
md/arch/xsimd_scalar.hpp:30:11:
note: template is dec
lared here
class batch;
^
3 errors generated.
---
I'm also attaching a workaround that allowed me to get past this, but it just
force-disables any SSE* usage on i386, so that's probably not the best
approach.