In D6479#96318, @pulkit wrote:

Since you did the nice work of moving shelve to core, how about rebasing this patch on tip of default with shelve in core?

In D6553#96119, @yuja wrote:

Need to update extensions._builtin so the shelve "extension" is silently
ignored.
Please send a follow up.

@pulkit Changes have been made.

@Kwan Thanks for the review. I have updated the patch as you suggested.

Abandoning since changes are reflected by the commit 9d31581cc44e.

Now, we have mergestate information storage in changeset level too with the help of extra mapping.

In D6448#94994, @mharbison72 wrote:

This needs some fixups to keep test-help.t and the Windows installer happy:
https://buildbot.mercurial-scm.org/builders/Win7%20x86_64%20hg%20tests/builds/1468/steps/run-tests.py%20%28python%202.7.13%29/logs/stdio
See 876f08f30ade for an example of how to fix the installer.

In D6479#94711, @pulkit wrote:

Can you fold both the commits, it will be better to see code for both storing and restoring mergestates in one patch.

Adding changes to D6479 and abandoning.

In D6479#94556, @pulkit wrote:

In D6479#94555, @navaneeth.suresh wrote:

Can we prevent merging again? Since we are getting content from old unresolved commit, copying mergestate from previously stored one, what purpose does this merging serves?

The usual unshelve mechanism restores the content of the files in the mergestate. However, we'll be still in a state with only one parent. This merge will help to recreate the exact same state with two parents.

There are other ways to set the second parent, like dirstate.setparents() I think. It is worth investigating to find a way to do it without merging again.

Can we prevent merging again? Since we are getting content from old unresolved commit, copying mergestate from previously stored one, what purpose does this merging serves?

Thanks for queuing Yuya.

I saw some patches[0][1][2] trying to fix this issue. It seems like it's difficult to preserve the second parent.

It doesn't make sense to check the existence of 'shelvedstate' in a static
table.
Can't we somehow get around the mq in a similar way to shelve.getcommitfunc()?
I don't think it's good idea to rely on state files saved on disk.

Is it difficult to fix unshelve to not lose merge parents?
https://bz.mercurial-scm.org/show_bug.cgi?id=5123#c2

Gentle ping for review.

Gentle ping for review.