How to merge to get rid of head with Mercurial command line, like I can do with TortoiseHg

branchmercurialmerge

My question is this:

  • If I have two heads (branches with changes) in my Mercurial repository, and I'd like to get rid of one of them, but discard all the changes from that branch instead of merging them into the other, and I can't strip out those changesets so I have to merge, how can I do that with the command line client?

If I have two heads in my Mercurial repository, and use TortoiseHg as my client, the repository might look like this:

two heads

Then I can get rid of the test2 head by doing a merge and discarding. First I would update to the head I'd like to keep (test3 in this case, which in the image above is already the current parent of my working folder). Then I would right-click and select "Merge with…":

merge with...

and in the dialog that pops up I would choose to discard the changes from the merge target (ie. the branch I'd like to discard all the changes from):

merge dialog

After this merge has gone through, all the changes in the test2 head has been discarded, and I can commit. The head has now disappeared, but the changeset is still part of history.

My question is this: How can I do the same thing using only the command line client? I can't find any matching options to the hg merge command:

hg merge [-P] [-f] [[-r] REV]

merge working directory with another revision

... snipped text

options:

 -f --force       force a merge with outstanding changes
 -t --tool VALUE  specify merge tool
 -r --rev REV     revision to merge
 -P --preview     review revisions to merge (no merge is performed)
    --mq          operate on patch repository

use "hg -v help merge" to show global options

Edit: debugsetparents worked nicely:

hg debugsetparents . 1
hg commit -m "merged to get rid of changeset #1"


Edit: Attempt to use the --tool internal:local according to one of the answers:

@echo off

setlocal
if exist repo rd /s /q repo
hg init repo
cd repo

rem revision 0
echo >test1.txt
hg commit -m "test1" --addremove

rem revision 1
echo >test2.txt
hg commit -m "test2" --addremove

rem revision 2
hg update 0
echo >test3.txt
hg commit -m "test3" --addremove

rem now let's get rid of change in revision 1 with merge
hg merge --tool internal:local -r 1
hg commit -m "merged"

dir

output of execution:

[C:\Temp] :test
adding test1.txt
adding test2.txt
0 files updated, 0 files merged, 1 files removed, 0 files unresolved
adding test3.txt
created new head
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
(branch merge, don't forget to commit)

 Volume in drive C is unlabeled      Serial number is 0e17:6aba
 Directory of  C:\Temp\repo\*

16.11.2010  20:05             .
16.11.2010  20:05             ..
16.11.2010  20:05             .hg
16.11.2010  20:05              13  test1.txt
16.11.2010  20:05              13  test2.txt
16.11.2010  20:05              13  test3.txt
                39 bytes in 3 files and 3 dirs    12 288 bytes allocated
    66 600 316 928 bytes free

Here the changes introduced in the 2nd changeset (the one with revision number 1), is now present in the merged changeset. This is not what i wanted.

Best Answer

According to TortoiseHG's source, when you check Discard all changes from merge target (other) revision, it uses the hg debugsetparents command:

hg debugsetparents REV1 [REV2]

manually set the parents of the current working directory

    This is useful for writing repository conversion tools, but should be used with care.

    Returns 0 on success.

use "hg -v help debugsetparents" to show global options

To use:

    hg up <revision-to-keep>
    hg debugsetparents <revision-to-keep> <revision-to-throw-away>
    hg commit -m "Merge to discard ..."