On Fri, Dec 31, 2021 at 09:19:34AM +0000, Stuart Henderson wrote:
> On 2021/12/31 06:51, Andrew Krasavin wrote:
> > Hello.
> >
> > This is another dependency for our telegram-desktop port (more
> > accurately, for tg_owt port). It seems more correct to get the shared
> > library as a separate port, that's why we refused to use the bundled
> > version.
> >
> > Keep in mind that Google supplies libyuv primarily as part of
> > Chromium and maintains build correctness only for the static library.
> > Unfortunately, it seems that these are not just words. For example,
> > one can observe this:
>
> Given what you've said about libyuv, I think this approach could be
> problematic and that it makes more sense to use the bundled copy.
We can always use the bundle and if libyuv means too many problems, we
certainly will.
We've split it out because it was doable without much effort.
I've deemed it worthwile because we can reduce the tg_owt/tdesktop
monolith by a noticable chunk, hopefully providing a port that will be
usable by other ports as well.
FWIW, FreeBSD also has a libyuv port, but it also seems unused at the
moment and their telegram ports still use the bundled version.
> As well as avoiding the problems you've mentioned, it reduces the
> risk of it interfering with build of the chromium-based ports
> (chromium/iridium/electron/qtwebengine) which won't expect a system copy
> of this library/headers.
chromium picking it up also occured to me but I haven't looked into
that yet.
Maybe robert knows more about this? I'm happy to beat this into shape
if it can actually benefit the browsers.
No comments:
Post a Comment