svn commit: r526863 - head/shells/ksh93-devel

[Jan Beich]

Cy Schubert <Cy.Schubert at cschubert.com> writes:

> In message <20200223180058.73ml7qthhj5phvn6 at atuin.in.mat.cc>, Mathieu 
> Arnold wr
> ites:
>
>> On Sun, Feb 23, 2020 at 09:02:34AM -0800, Cy Schubert wrote:
>> > In message <20200223154559.v4hlnjn24f7fxr2g at atuin.in.mat.cc>, Mathieu=20
>> > Arnold wr
>> > ites:
>> > > > >
>> > > > > How can the commit date change and the hash remain the same?  This =
>> is a
>> > > > > very very unlikely collision.
>> > > >=3D20
>> > > > That was purely my fault. I had built and tested the port and committ=
>> ed i=3D
>> > > t=3D20
>> > > > to a branch of my git tree. Prior to merging it to master I changed t=
>> he=3D
>> > > =3D20
>> > > > commit date a more recent date to reflect the git svn dcommit I was=
>> =3D20
>> > > > planning that day. Then did git rebase to redo the commit, git merge =
>> to=3D
>> > > =3D20
>> > > > merge back to master, and finally git svn dcommit. I missed a step in=
>>  my=3D
>> > > =3D20
>> > > > workfolow.
>> > >
>> > > I still do not understand, how does that influence the date the commit
>> > > referenced by the hash, 0be82553 here, was made?
>> >=20
>> > OK, here's the timeline.
>> >=20
>> > Feb 8:
>> >=20
>> > 1. I branch master to ksh93 in my github repo.
>> > 2. The port is updated. COMMIT_DATE=3D2020-02-08
>> > 3. make makesum
>> > 4. Port is built, builds correctly.
>> > 5. Port is committed to my ksh93 branch. (The other ksh93 is also=20
>> > committed.)
>> >=20
>> > Feb 22:
>> >=20
>> > 6. git merge from ksh93 to master
>> > 7. I realize the COMMIT_DATE of 2020-02-08 is now incorrect.
>> > 8. I change the commit date to the current date and use git rebase to=20
>> > re-commit the commit.
>> > 9. (I forgot to make maksum when I changed the commit date. My bad.)
>> > 10. git svn dcommit to push my commits to svn.
>> >=20
>> > That's how it happened. Do you understand now?
>>
>> Mmmm, ok, but, the version of the port is supposed to be about upstream,
>> not some made up date you create.  In this case, the version is supposed
>> to be the date of the *upstream* commit, not the date where you do
>> stuff.
>
> That makes sense. I started doing this when I realized github didn't report 
> commits accurately so I opted to use the date I did the work.

GitHub does report CommitDate in addition to AuthorDate e.g.,

https://github.com/krb5/krb5/commits/master

where "Commits on Feb 17, 2020" refers to CommitDate but if you open
the specific commit its AuthorDate is from "Nov 26, 2019".