zstd and Mercurial packaging

Augie Fackler raf at durin42.com
Thu Dec 1 22:31:46 UTC 2016


> On Dec 1, 2016, at 17:29, Sean Farley <sean at farley.io> wrote:
> 
> Augie Fackler <raf at durin42.com> writes:
> 
>>> On Dec 1, 2016, at 17:07, Sean Farley <sean at farley.io> wrote:
>>> 
>>> Augie Fackler <raf at durin42.com> writes:
>>> 
>>>>> On Dec 1, 2016, at 08:53, Neal Becker <ndbecker2 at gmail.com> wrote:
>>>>> 
>>>>> Fedora guidelines: All packages whose upstreams allow them to be built against system libraries must be built against system libraries.
>>>>> 
>>>>> So if zstd could become a standard system library it would have to be unbundled.
>>>> 
>>>> That should be possible some day, but today we're depending on an as-yet unstable API that requires static linking against a specific version of the library. Is that sufficient to prevent unbundling for now?
>>> 
>>> Well, that's problematic for me since I've already added zstd as a
>>> system lib in MacPorts (under the, I believe, fair assumption) that this
>>> would be easy to override and link against.
>> 
>> It won't hurt anything to have it there, it just won't help you until zstd stabilizes their API and we can get on the stable API.
>> 
>> It was a fair assumption, except for all the discussion both here and on indygreg's patches which made it clear that it wasn't ;)
> 
> I don't understand why there isn't a switch to enable this?

I'm not sure what "this" is - if "this" is "using zstd at all" then yes, I believe that's in there. If it's "using a system zstd instead of the vendored one" that's *technically infeasible* with the current stable zstd API. The wire format is stable, but the API is still under some amount of development and they're being careful.

> When I read
> the patches, I thought that was the goal? I'm a bit disappointed in this
> situation now and retract my support of vendoring zstd.

If it's that worrisome, have MacPorts ship hg without zstd - that'll always be fine.


More information about the Mercurial-packaging mailing list