[Webkit-unassigned] [Bug 197192] [GTK][WPE][JSCOnly] compile error when -DWTF_CPU_ARM64_CORTEXA53=ON set for arm64

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=197192

Carlos Alberto Lopez Perez <clopez at igalia.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |bugs-noreply at webkitgtk.org,
                   |                            |ticaiolima at gmail.com
           See Also|                            |https://bugs.webkit.org/sho
                   |                            |w_bug.cgi?id=138315
            Summary|compile error when          |[GTK][WPE][JSCOnly] compile
                   |-DWTF_CPU_ARM64_CORTEXA53=O |error when
                   |N set for arm64             |-DWTF_CPU_ARM64_CORTEXA53=O
                   |                            |N set for arm64

--- Comment #2 from Carlos Alberto Lopez Perez <clopez at igalia.com> ---
'-DWTF_CPU_ARM64_CORTEXA53=ON' Enables a workaround for a CPU Bug (erratum 835769) present in early versions of ARM64 Cortex A53 (in more detail, according to what I find in Internet, to revisions <= r0p4 of the Cortex A53 CPU).

This workaround was added 5 years ago in r175514 but it seems it stopped building after r236589 removed removes function JSC::AssemblerBuffer::data() for ARM64.

There are two major problems with this workaround:

1) The way it gets enabled its completely broken for cross-building. The CMake logic tries to parse /proc/cpuinfo to see if the build machine is a Cortex A53 (it just assumes build_machine==target_machine). It also doesn't do any extra effort to match the revision of the CPU to see if its newer than r0p4. It simply enables it for any build_machine==cortex_a53.

2) Testing that the workaround continues to be necessary and that it continues to work is tricky because it requires the affected hardware available.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20191008/1f6d0261/attachment-0001.html>