Re: merge from emacs-25 to master lost a line

[Phillip Lord]

Paul Eggert <address@hidden> writes:

> The recent merge from emacs-25 to master lost the attached one-line spelling
> fix for test/automated/url-parse-tests.el that was originally installed in
> emacs-25 commit c9dfe5121cb31aa1e8cf352e5088940fc5dd7106 dated 2015-12-26.
> This fix was lost even though the commit and change is listed in the master
> branch now. I fixed the problem in master as part of commit
> 61e83e902b388490b609677a76f3d49740439f24 dated 2016-01-04.
>
> Although this particular bug is both minor and fixed now, I worry that there
> may be similar problems elsewhere.
>
> I suspect the problem arose due to the big renaming of tests in commit
> 22bbf7ca22f11cc33d887d0162cf2ec6661c3a3e dated 2015-11-23, which renamed a
> bunch of test files, e.g., from test/automated/url-util-tests.el to
> test/lisp/url/url-util-tests.el. It looks like this renaming means that if
> emacs-25 adds a test (under the old naming scheme) and then later changes the
> test (also under the old naming scheme), the later change will be lost when be
> merging to master.

Hmmm. I was hoping that git was clever enough to understand the file
move, or that, at least, this would produce a conflict,
url-util-tests.el was modified on one branch and deleted (i.e. moved) in
another.


> If my analysis is correct, I suggest two things:
>
> * First, when merging to master we have to be careful about the renamed tests.
> I suppose somone (I hope it's not me) could go through all the emacs-25
> changes to tests renamed in master, and check that the changes are properly
> merged to master.

As this is my fault, I will work on this, or at least start to. It
should at least allow me to understand if this is just url-util-tests.el
or general problem as you fear.


> * Second, to help avoid this problem in the future we should avoid making
> wholesale renames to master right now. They're not worth the hassle. That is,
> it's OK to add stuff to master, but let's not rename files in master unless we
> really need to.

Yes, it looks like I should have waited till after the emacs-25 release
was made.

Oh dear, I was ever over-enthusiastic.

Phil