Re: continuing firefox woes: wrong USE_LANGUAGE, missing GCC_REQD

[Greg Troxel <gdt%lexort.com@localhost>]

>> Looking in mozilla-common.mk, I find this:
>>
>>   # Some modules written in Rust dislike gnu++17 as of 74.0.
>>   USE_LANGUAGES+=         c99 c++
>>
>> I won't understand how this can be correct; I have heard that firefox
>> requires C++17, and the above declares that C++03 is enough.
>>
>> Even if there is some issue with 17, shouldn't that say c++14?
>>
>> Also, there is no GCC_REQD.  NetBSD 8 -- squarely on the supported list
>> of pkgsrc -- has gcc5, and I have the impression that it is known to
>> fail.
>
> www/firefox uses clang++ as C++ compiler.
> I think that GCC_REQD is not relevant to www/firefox.

It would be really useful to explain that in a commend near
USE_LANGUAGES.

In my work/ tree, I have .gcc/bin and no .clang/bin.  So at least the
pkgsrc infrastructure thinks that gcc is being used.  But I see that
clang is being used, and it's not clear that this is through the
wrapppers.  It would be nice to have that explained in
mozilla-common.mk.

> Some parts in www/firefox require gnu++17, and some parts
> require gnu++14 and they dislike gnu++17.
> www/firefox cannot accept consistent -std=... flags.
> Adding -std=gnu++14 or -std=gnu++17 to all c++ invocations
> causes build failures.

It sounds like firefox adds the flags it needs.

Perhaps pkgsrc should be changed to separate "compiler must be capable
of" from "pkgsrc wrappers will add --std".   In my view, it's a bug if
a build system doesn't pass the --std that is appropriate for each file
that it builds.

> If anyone has proper way to handle this situation of www/firefox,
> feel free to fix this problem.
>
> And I will find a time to fix the build on NetBSD/{i386,amd64} 8.

My build of 75 failed in the same way that 


Thanks.  I think it would be really nice to help our official 2020Q1
builds have a working firefox.