Re: firefox build failure on NetBSD 8.0

[Chavdar Ivanov <ci4ic4%gmail.com@localhost>]

Firefox 73.0 also just failed build for me under -current (9.99.47
from a few days ago, amd64):
....
[style 0.0.1] /usr/pkgsrc/www/firefox/work/build/dist/include/mozilla/cxxalloc.h:59:1:
warning: replacement function 'operator delete[]' cannot be declared
'i[1153/1913$nline-new-delete]
[style 0.0.1] /usr/pkgsrc/www/firefox/work/build/dist/include/mozilla/cxxalloc.h:63:1:
warning: replacement function 'operator delete[]' cannot be declared
'inline' [-Wi
nline-new-delete]
[style 0.0.1] /usr/pkgsrc/www/firefox/work/build/dist/include/mozilla/RefPtr.h:116:46:
error: no member named 'is_convertible_v' in namespace 'std'
[style 0.0.1] /usr/pkgsrc/www/firefox/work/build/dist/include/mozilla/RefPtr.h:116:63:
error: 'I' does not refer to a value
[style 0.0.1] /usr/pkgsrc/www/firefox/work/build/dist/include/mozilla/RefPtr.h:115:22:
note: declared here
[style 0.0.1] /usr/pkgsrc/www/firefox/work/build/dist/include/mozilla/RefPtr.h:116:65:
error: expected expression
[style 0.0.1] /usr/pkgsrc/www/firefox/work/build/dist/include/mozilla/RefPtr.h:117:16:
error: expected ',' or '>' in template-parameter-list
[style 0.0.1] /usr/pkgsrc/www/firefox/work/build/dist/include/mozilla/RefPtr.h:122:23:
error: expected member name or ';' after declaration specifiers
[style 0.0.1] /usr/pkgsrc/www/firefox/work/build/dist/include/mozilla/gfx/2D.h:1128:27:
error: no viable conversion from
'already_AddRefed<mozilla::gfx::SourceSurface>'
to 'RefPtr<mozilla::gfx::SourceSurface>'
....


seems to be the firs error; at the end one sees:
....
, "/usr/pkgsrc/www/firefox/work/build/dist/include/mozilla/ServoTraversalStatistics.h",
"/usr/pkgsrc/www/firefox/work/build/dist/include/mozilla/SizeOfState.h",
"/usr/pk
gsrc/www/firefox/work/build/dist/include/nsCSSProps.h",
"/usr/pkgsrc/www/firefox/work/build/dist/include/nsContentUtils.h",
"/usr/pkgsrc/www/firefox/work/build/dist/incl
ude/nsNameSpaceManager.h",
"/usr/pkgsrc/www/firefox/work/build/dist/include/nsMediaFeatures.h"]',
servo/components/style/build_gecko.rs:195:13
stack backtrace:
   0:        0x18c80d1e3 -
<std::sys_common::backtrace::_print::DisplayBacktrace as
core::fmt::Display>::fmt::h3e9ed830bd4974ee
   1:        0x18c824e0d - core::fmt::write::h635c1c51677f3dab
   2:        0x18c80bf05 - std::io::Write::write_fmt::h03284227df09b675
   3:        0x18c7fce4e -
std::panicking::default_hook::{{closure}}::heb424b4be6f41adb
   4:        0x18c7fcb42 - std::panicking::default_hook::h8d750f9d956cec16
   5:        0x18c7fd52e -
std::panicking::rust_panic_with_hook::h2577a268bff4094b
   6:        0x18c7fd0ce - std::panicking::continue_panic_fmt::h60ea77b3990a4827
   7:        0x18c7fd00f - std::panicking::begin_panic_fmt::h7461673961d6ac46
   8:        0x18c64ebfb -
build_script_build::build_gecko::generate_structs::h9c1b46512edf0288
   9:        0x18c65079e - build_script_build::main::h979df466336aff00
  10:        0x18c637383 - std::rt::lang_start::{{closure}}::h6f88d729e6eb10b1
  11:        0x18c7fcf63 - std::panicking::try::do_call::h7ccd7d4960560c1c
  12:        0x18c810eea - __rust_maybe_catch_panic
  13:        0x18c8023f9 - std::rt::lang_start_internal::h70488ea29cda7a04
  14:        0x18c6511d2 - main
  15:        0x18c63474b - ___start
....


On Sat, 15 Feb 2020 at 18:19, Tobias Nygren <tnn%netbsd.org@localhost> wrote:
>
> On Sat, 15 Feb 2020 20:45:02 +0530
> Mayuresh <mayuresh%acm.org@localhost> wrote:
>
> > Is it one of those things that won't build on 8.0? Or is it something that
> > requires compiler upgrade? And if it's the latter what's a systematic way
> > to do so, so that things work out of the box for 8.x users?
>
> The actual error message will be further up (can be quite far up with
> MAKE_JOBS>1). It does need C++17 now and yes, the C++ headers on 8.x are
> too old.
>
> Normally adding "gnu++17" to USE_LANGUAGES (which I have just done)
> would ensure that a recent enough gcc (and libstdc++ + headers) would
> be used. but since Firefox insists on being built with clang there is
> no such logic and we lose.
>
> Properly fixing this will be non-trivial. A workaround might be to patch
> the package to have a local copy of the necessary parts of the
> "g++/type_traits" header file that are added with C++17.
>
> Kind regards,
> -Tobias



-- 
----