[swift-build-dev] [swift-evolution] Proposal: Package Manager Version Pinning

Alexis Beingessner abeingessner at apple.com
Fri Oct 14 19:17:55 CDT 2016



> On Oct 14, 2016, at 8:00 PM, Paul Cantrell <cantrell at pobox.com> wrote:
> 
> 
>>> On Oct 14, 2016, at 6:42 PM, Daniel Dunbar <daniel_dunbar at apple.com> wrote:
>>> 
>>> On Oct 14, 2016, at 4:02 PM, Paul Cantrell <cantrell at pobox.com> wrote:
>>> 
>>> I’m puzzled. If a package’s pinning does not affect any other package that uses it, why should the defaults be different? A library will still suffer from all the “works for me” problems an app might.
>>> 
>>> Is the rationale that not pinning libraries encourages accidental testing of new versions of a library’s dependencies as they arrive? Or is there another rationale for having different defaults?
>> 
>> I'll defer to this comment (linked from someone else earlier in the thread), which happens to match up with my perspective:
>> https://github.com/yarnpkg/yarn/issues/838#issuecomment-253362537
> 
> I took that comment to be an explanation of why a library's lockfile/pinfile should not propagate to other packages that use it. That is clearly the case; such pin propagation would be nonsensical.
> 
> My question was not about that, but about why libraries shouldn’t use a pinfile at all, even for their own _internal_ development. All the same “last know good build” concerns apply.
> 
> The difference is that testing against that single last known good version set is sufficient for a top-level package, whereas a library should (1) ideally test against multiple valid dependency versions and (2) test often against new versions of its dependencies. I don’t see, however, that this implies that libraries should not have pinfiles at all — just that their release / CI process should not be limited to what’s pinned.

A few comments down, Yehuda even provides an example of him doing just that with Bundler:

https://github.com/yarnpkg/yarn/issues/838#issuecomment-253366352

But in this case you actually want to maintain *many* lock files, and so it seems fine to require a bit of extra work (passing some flags) to do this. Drifting tests are the better default here. It makes library CI into an alpha-tester, empowering binaries to be more confident in upgrading frequently.

> 
> Cheers, P
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.swift.org/pipermail/swift-build-dev/attachments/20161014/26718222/attachment.html>


More information about the swift-build-dev mailing list