Re: Git p4 sync changelist interval

[Luke Diamand <luke@diamand.org>]

On 6 June 2017 at 08:00, Luke Diamand <luke@diamand.org> wrote:
> On 6 June 2017 at 00:29, Luke Diamand <luke@diamand.org> wrote:
>> On 5 June 2017 at 19:50, Андрей Ефанов <1134togo@gmail.com> wrote:
>>> 2017-06-04 14:09 GMT+03:00 Luke Diamand <luke@diamand.org>:
>>>>
>>>>
>>>> >
>>>> > The problem, as I see it, is that before syncing changes in the given
>>>> > range, p4 task does not sync to cl1 version of the repo, and applies
>>>> > commits to the empty repository.
>>>> > Is it a bug or my misunderstanding of how git p4 should work?
>>>>
>>>> Possibly I'm misunderstanding what you're doing! Can you give a
>>>> sequence of steps to show the problem?
>>>
>>> What I meant is:
>>>
>>> 1. Create p4 depot
>>> 2. Add first.file (CL 2)
>>> 3. Add second.file (at CL 3)
>>> 4. Add third.file (at CL 4)
>>> 5. Modify first.file (at CL 5)
>>> 4. git-p4 clone //depot@3,5
>>>
>>> In this case first.file, will not be represented in the repository.
>>
>> Hmmm, it's not working right for me. Although in my case I seem to be
>> missing the second file.
>>
>> It's fine if I don't use the revision range "3,5".
>
> It's also fine if I do:
>
> git p4 sync //depot@3
> cd depot
> git p4 rebase

It seems to be something to do with the code around line 3395 that says:

    if self.changeRange == "@all":
        self.changeRange = ""
    elif ',' not in self.changeRange:

It's finding a lower revision number with which to later call
importHeadRevision(), but that only seems to be called if the revision
range does *not* have a "," in it. As a result, we never actually call
importHeadRevision() and so files end up missing.

The obvious fix of fishing out the "@3" from the "@3,5" revision range
works in this instance, but breaks some of the regression tests.

Luke