Re: CVS commit: pkgsrc

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

Leonardo Taccari <leot%NetBSD.org@localhost> writes:

> Hello Ryo,
>
> Ryo ONODERA writes:
>> [...]
>> Log Message:
>> Add graphite2 to librsvg via pango and harfbuzz
>>
>> librsvg with pango and harfbuzz with graphite2 is required for libreoffice.
>> librsvg-c has the problem on NetBSD and libreoffice uses internal static
>> harfbuzz library. Rust-based librsvg has no problem and libreoffice use
>> dynamic harfbuzz from fonts/harfbuzz.
>>
>> Bump PKGREVISIONs.
>> [...]
>
> Shouldn't this check be made based on LIBRSVG_USE_RUST variable
> (instead of PLATFORM_SUPPORTS_RUST)? (Otherwise possible user's
> preferences about librsvg can interfere with that.)
>
> I still haven't looked at it in details but I think that the current
> LIBRSVG_USE_RUST is limited to buildlink3 logic while librsvg is
> often used as a tool dependency too (for rsvg-convert) and like in
> this case only the LIBRSVG_USE_RUST value is needed.  I think it's
> probably worth to separate the logic il graphics/librsvg/buildlink3.mk
> in two new files:
>
>  - available.mk: to just define and handle LIBRSVG_USE_RUST variable
>  - tool.mk:      to make librsvg as a tool dependency (ATM all
>                  BUILD_DEPENDS and TOOL_DEPENDS hardcodes
>                  graphics/librsvg ignoring LIBRSVG_USE_RUST.  AFAICS the
>                  only exception is x11/mate-desktop that despite needing
>                  librsvg as a tool dependency picks it up via bl3 so
>                  LIBRSVG_USE_RUST is honored)
>
> If noone beat me I hope to discuss a possible patch ASAP in tech-pkg@.

I think you are right that this needs to be a bit more complicated.

In particular, I think we should support the notion of someone choosing
to avoid rust on a platform where it theoretically works (because their
computer is slow, because they don't like it, or because it doesn't
actually work).