Bug#978098: webkit2gtk: Re-enable build for hurd-i386?

[Alberto Garcia]

On Mon, Dec 28, 2020 at 09:01:21PM +0100, Samuel Thibault wrote:

> > Shouldn't the patch be applied in "official" package as well (and
> > maybe upstream)?
> 
> Debian maintainers usually prefer to see patches applied upstream
> before adding them to their packaging. They are already submitted
> upstream, see URLs in the patch file. Some of them are already
> applied, others need some rework, help welcome!
> 
> And they usually rather see a complete working set before changing
> debian/

Hi Laurent and Samuel,

I'm generally happy to make my packages work in as many architectures
as possible, and that includes the non-Linux ports.

WebKit is however a bit special because of its size, complexity and
fast development pace. We often find problems with architectures other
than the "main" ones (intel, arm) and it sometimes takes a while to
figure out what's going on. It's even harder with the Hurd or kFreeBSD
ports. Because of that I decided to stop building WebKit packages for
those systems some time ago.

However if someone has patches and can verify that the package builds
and works fine with them I'm happy to include them.

I had a quick look at this patch in particular and it seems that at
least parts of it could go upstream just fine, and I can try to help
wit that if you want. I'm happy to generally reduce or remove the
dependency on PATH_MAX in WebKit.

I have a question that maybe you can answer:

> ++#if _POSIX_VERSION >= 200112
> ++        char *normalizedPath;
> ++        normalizedPath =
> realpath(FileSystem::fileSystemRepresentation(filePath).data(),
> NULL);
> ++        if (!normalizedPath)
> ++            continue;
> ++
> ++        filePath =
> FileSystem::stringFromFileSystemRepresentation(normalizedPath);
> ++        free(normalizedPath);
> ++#else

I'm not familiar with the _POSIX_VERSION macro, but realpath(3) says
that passing NULL to that function is implementation-defined in
POSIX.1-2001, and only properly specified in POSIX.1-2008, so I would
expect that you need to check for a different version.

Either way, is there any modern system that does not implement
POSIX.1-2008?

   -----

I also saw that Samuel already sent the changes to computeRAMSize()
upstream, and that earlier someone added support for FreeBSD, so maybe
we can enable replace 'linux-any hurd-any' by simply 'any'.

Regards,

Berto