tech-pkg archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: CVS commit: pkgsrc/graphics/librsvg



Joerg Sonnenberger <joerg%bec.de@localhost> writes:

> On Sat, Dec 26, 2020 at 11:39:30PM +0000, Nia Alarie wrote:
>> Module Name:	pkgsrc
>> Committed By:	nia
>> Date:		Sat Dec 26 23:39:30 UTC 2020
>> 
>> Modified Files:
>> 	pkgsrc/graphics/librsvg: Makefile available.mk
>> 
>> Log Message:
>> librsvg: Restore default of librsvg-c on ARM
>> 
>> Avoid rust version in pbulk when C version is selected
>> 
>> Follows comments by gdt by email
>
> I'm really, really annoyed by this kind of change. It is technically
> wrong, the comment is complete BS and it breaks things for no good reason
> and it starts from a lack of debugging. Why do we have to rush this?

So let's discuss the actual issue and solutions.

First, we used to default to librsvg-c, because rust doesn't reliably
build on typical armv7 machines.  So that part is just getting back to
where we were.

We have a real problem with bulk builds, where the rust version succeeds
(because the bulk machine is very beefy) and then it gets installed
instead.

We need a near-zero-risk in a small number of days, which ends up
trading off risk and ick.  nia@ suggested, and I agreed, that just not
building librsvg (rust) in bulk builds where librsvg-c is selected, was
a decent workaround for this branch, to result in users being able to
build things and bulk builds building packages that depend on librsvg.

If you think we should do something different -- which apparently you
do, but you only said you didn't like this, not what should be done --
then please mail a proposed diff, or precise enough description to
understand, to tech-pkg, explain why it's right, and include rationale
for why it is near-zero risk of trouble for 2020Q4.  I am more than
willing to read a diff, ask others to read it, and listen to the
reasoning.


Attachment: signature.asc
Description: PGP signature



Home | Main Index | Thread Index | Old Index