vcs_info backports? (was: test release: 5.9.1.2-test (will be 5.9.2))

dana wrote on Sun, 28 Jun 2026 20:52 +00:00:
> if you have time, please test and report any issues

% git log --oneline zsh-5.9..origin/master Functions/VCS_Info/ | me --ex=
clude-already-cherry-picked
510df60dd 51144, 51146: vcs_info git: stg: Extract patch descriptions
7cdada166 51142: vcs_info git: Check the get-unapplied style as document=
ed
ca7c42e1e 51138: Updated StGit patch detection in vcs_info
%=20

workers/51138: Does this patch meet our criteria for backporting to a pa=
tch release?  It replaces an implementation that only works with older s=
tg(1) versions with an implementation that works with both old and new s=
tg(1) versions.  The underlying problem was reported to StGit as a regre=
ssion back in 2022: <https://github.com/stacked-git/stgit/issues/183>. =20

workers/51142 seems like it could easily break people's setups ("I upgra=
ded and vcs_info stopped showing information from git-rebase(1)'s todo l=
ist"), so not proposing it.  If anything, it should perhaps be mentioned=
 in 5.10's list of changes in README?

workers/51144+51146... Not sure.  Smoke testing the --description flag a=
gainst 2019's stg(1) produces the same output format as in current stg(1=
), so I suspect the new code would work correctly even against the versi=
ons of stg(1) zsh-5.9.next is likely to be run against.  As to potential=
 of regressions: The change only affects people who use stg(1) and have =
a gen-applied-string/gen-unapplied-string hook =E2=80=94 and the thing i=
s, people who use stg(1) very likely have run into conflicts during reba=
ses at some point (that's part and parcel of maintaining a patch series)=
, which would have exercised those same hooks of theirs on the git-rebas=
e version of gen-applied-string/gen-unapplied-string data, which contain=
s patch descriptions.  Therefore, I suspect backporting this is more lik=
ely to cause existing hooks to start DTRTing in more cases than to cause=
 a regression =E2=80=94 but on the other hand, it's not bugfix but new f=
unctionality, and I'm not aware of a user reporting the lack of patch su=
bject lines extraction as a bug / feature request.

If I've missed the 5.9.2 release train then please consider the question=
s for 5.9.3, should there be one (even if not soon).

Thanks,

Daniel