[oe] chromium detected as stripped

[Raphael Kubo da Costa]

Hey Trevor,

Thanks again for testing the recipe and providing great feedback.

Trevor Woerner <twoerner at gmail.com> writes:
> When I compile chromium61 with pyro, I get a strange error. Every time the
> python wrapper (gcc_link_wrapper.py) is called for a native compile (g++), I
> get:
[...]
> This happens with both qemux86-64 and raspberrypi3. Unfortunately those log
> files are 18M and 19M each, so I won't be attaching them. Why can't my build
> find 'ld'? That seems quite strange. Probably some kind of ld vs gold issue?
> I'm not using gold anywhere and have not enabled it, are you using gold?

You are correct. By default, Chromium will default to using gold (and
passing -fuse-ld=gold like it did above) in most situations for both
host and target builds (the error you posted is from building a native
tool), as that's what upstream uses and tests.

I've changed my recipe to respect the ld-is-gold DISTRO_FEATURE; the
downside is that this setting applies to both host and target, so if one
decides to use gold it also needs to be present on the host system:
<https://github.com/rakuco/meta-crosswalk/commit/e7cfbf616f0bc524f9487d98c6e39e30ef28a639>

> Compiling chromium61 with master gives a long error in breakpad. That log file
> is 73K so I've attached it (log.do_compile__qemux86-64__chromium61_master_breakpad).
> It appears to be a problem with the code itself. Here is the start:
[...]
> 	../../breakpad/src/client/linux/dump_writer_common/ucontext_reader.cc: In static member function 'static uintptr_t google_breakpad::UContextReader::GetStackPointer(const google_breakpad::ucontext*)':
> 	../../breakpad/src/client/linux/dump_writer_common/ucontext_reader.cc:92:12: error: invalid use of incomplete type 'const struct google_breakpad::ucontext'

This is actually caused by glibc 2.26, which removed 'struct ucontext'
in favor of ucontext_t
https://sourceware.org/glibc/wiki/Release/2.26#Removal_of_.27struct_ucontext.27

meta-browser was fixed by Khem Raj with commit c2e44f94636 ("chromium:
Fix build with security flags"), but my recipe didn't have anything
similar.

It turns out upstream actually fixed this issue a few days ago, so I was
able to just backport their fix here:
<https://github.com/rakuco/meta-crosswalk/commit/e7cfbf616f0bc524f9487d98c6e39e30ef28a639>