unclear test-merge5 and update -c

[Stuart W. Marks]

Gilles Moris wrote:
> This is related to a series of patch from Stuart Marks one month ago, 
> about being able
> to update without option flag from a clean work tree:
> http://www.selenic.com/pipermail/mercurial-devel/2009-August/014835.html
> 
>  From the history, I think this was possible before cdc458b12f0f, but 
> Matt then
> changed the logic, but did not update the comment in the output, so that 
> the test
> is correct with the current logic (no update crossing heads of the same 
> branch),
> but the comment is wrong.

Yes, I had noticed this issue with the mismatch between the comment and the 
actual behavior in test-merge5. It turns out that this will be "fixed" by my 
patch. Where the comment says "should succeed" it will actually succeed, 
whereas it currently fails.

> I would like Matt and the crew to look again at the inclusion of Stuart 
> patch series
> to get update from a clean work tree between any heads working without 
> any -c or -C.
> 
> Then, the last update in test-merge5 will succeed again.

Yes, that's right. Thanks for your support. I've updated the patch per your 
comments from the last time I posted them, and I'll repost them shortly 
(assuming all tests pass).

s'marks

> 
> Regards.
> Gilles.
> 
> On Wed, Sep 16, 2009 at 1:07 AM, Adrian Buehlmann <adrian at cadifra.com 
> <mailto:adrian at cadifra.com>> wrote:
> 
>     Hi Matt and list
> 
>     Today I ran across the problem that I had to specify -c in
>     order to be able to update from one head to another.
> 
>     I noticed that the error message I get without -c seems
>     not quite correct, as mercurial recommends to use -C (big)
>     even though -c (small) would do fine.
> 
> 
>     So I was trying to prepare the following patch:
> 
>     <insert>
>     merge: do not recommend to use -C if -c would do when aborting
> 
>     diff --git a/mercurial/merge.py b/mercurial/merge.py
>     --- a/mercurial/merge.py
>     +++ b/mercurial/merge.py
>     @@ -444,7 +444,7 @@
>                         raise util.Abort(_("crosses branches (use 'hg
>     merge' or "
>                                            "'hg update -C' to discard
>     changes)"))
>                     raise util.Abort(_("crosses branches (use 'hg merge' "
>     -                                   "or 'hg update -C')"))
>     +                                   "or 'hg update -c')"))
>                 elif wc.files() or wc.deleted():
>                     raise util.Abort(_("crosses named branches (use "
>                                        "'hg update -C' to discard
>     changes)"))
>     </insert>
> 
> 
>     Then I ran the testsuite and came across test-merge5, which in
>     my book has as a rather questionable history.
> 
> 
>     In 6aa4a3fa4e60, Matt changed test-merge5 as follows:
> 
>     <insert>
>     update: no -C needed to switch branches without outstanding changes
> 
>     diff --git a/tests/test-merge5 b/tests/test-merge5
>     --- a/tests/test-merge5
>     +++ b/tests/test-merge5
>     @@ -12,9 +12,14 @@ hg commit -m"comment #1" -d "1000000 0"
>      hg update 0
>      rm b
>      hg commit -A -m"comment #2" -d "1000000 0"
>     +mv a c
>      # in theory, we shouldn't need the "-y" below, but it prevents
>      # this test from hanging when "hg update" erroneously prompts the
>      # user for "keep or delete"
>     +echo % should abort
>     +hg update -y 1
>     +mv c a
>     +echo % should succeed
>      hg update -y 1
> 
>      exit 0
>     diff --git a/tests/test-merge5.out b/tests/test-merge5.out
>     --- a/tests/test-merge5.out
>     +++ b/tests/test-merge5.out
>     @@ -1,4 +1,7 @@
>      1 files updated, 0 files merged, 0 files removed, 0 files unresolved
>      removing b
>      created new head
>     +% should abort
>      abort: update spans branches, use 'hg merge' or 'hg update -C' to
>     lose changes
>     +% should succeed
>     +1 files updated, 0 files merged, 0 files removed, 0 files unresolved
>     </insert>
> 
> 
>     So that change specifically introduced a test to check that update
>     succeeds
>     without specifying -C in that case.
> 
> 
>     But in later cdc458b12f0f, Matt changed test-merge5.out as follows:
> 
>     <insert>
>     diff --git a/tests/test-merge5.out b/tests/test-merge5.out
>     --- a/tests/test-merge5.out
>     +++ b/tests/test-merge5.out
>     @@ -2,6 +2,6 @@ 1 files updated, 0 files merged, 0 files
>      removing b
>      created new head
>      % should abort
>     -abort: update spans branches, use 'hg merge' or 'hg update -C' to
>     lose changes
>     +abort: crosses branches (use 'hg merge' or 'hg update -C' to
>     discard changes)
>      % should succeed
>     -1 files updated, 0 files merged, 0 files removed, 0 files unresolved
>     +abort: crosses branches (use 'hg merge'or 'hg update -C')
>     </insert>
> 
> 
>     Notice that the "should succeed" does *not* succeed: it produces
>     an abort, and thus exactly fails the test that was introduced
>     in 6aa4a3fa4e60.
> 
> 
>     Currently we have the new-ish option -c which seems to help exactly
>     in that case, namely allow an update to another head in the
>     same named branch if there are no local modifications.
> 
> 
>     Question is: is test-merge5 and test-merge5.out correct?
> 
>     In other words: should test-merge5 be updated to match its output,
>     or is the output wrong?
>     _______________________________________________
>     Mercurial-devel mailing list
>     Mercurial-devel at selenic.com <mailto:Mercurial-devel at selenic.com>
>     http://selenic.com/mailman/listinfo/mercurial-devel
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Mercurial-devel mailing list
> Mercurial-devel at selenic.com
> http://selenic.com/mailman/listinfo/mercurial-devel