Closed Bug 1553068 Opened 6 years ago Closed 6 years ago

A few hashes missing in the git -> hg map

Categories

(Developer Services :: Git, defect)

defect
Not set
normal

Tracking

(Not tracked)

RESOLVED WONTFIX

People

(Reporter: marco, Assigned: dhouse)

References

Details

Attachments

(1 file)

Attached file missing.txt

I'm attaching a file with a list of git hashes missing (each line corresponds to a hash).

Dave, can you force the remapping?

Flags: needinfo?(dhouse)
Assignee: nobody → dhouse
Flags: needinfo?(dhouse)

none of the commits were found in the local mapfile:

$ cat missing.txt | xargs -I {} grep {} ./build/conversion/beagle/.hg/git-mapfile
$ 

but the commits are present in the local git repo

I verified none are found in the mapper:

$ cat missing.txt | xargs -P15 -I {} bash -c 'curl -s --fail -X GET --header "Accept: application/json" "https://mapper.mozilla-releng.net/gecko-dev/rev/hg/{}"; exit 0' | tee found-missing.txt
$ 

oh, these appear to be old commits. the ones I've viewed are from 2005
So, I may need to manually create the mapping.

Could you create the mapping for all commits? Or would that be expensive?
I'm doing some analyses on the history, so I'd need many of the old commits to be mapped.

(In reply to Marco Castelluccio [:marco] from comment #3)

Could you create the mapping for all commits? Or would that be expensive?
I'm doing some analyses on the history, so I'd need many of the old commits to be mapped.

I think we can create the mappings for all commits. There are a few waypoints where the conversion changed, and so it will take some time to work through the history.

When do you need the mappings for all commits? I will start working on the mapping between things, but unless you need it sooner I think it could be about a week (as I fit in time working on it between other tasks).

Flags: needinfo?(mcastelluccio)

(In reply to Dave House [:dhouse] from comment #4)

(In reply to Marco Castelluccio [:marco] from comment #3)

Could you create the mapping for all commits? Or would that be expensive?
I'm doing some analyses on the history, so I'd need many of the old commits to be mapped.

I think we can create the mappings for all commits. There are a few waypoints where the conversion changed, and so it will take some time to work through the history.

When do you need the mappings for all commits? I will start working on the mapping between things, but unless you need it sooner I think it could be about a week (as I fit in time working on it between other tasks).

Ideally as soon as you can, as it is blocking me unfortunately :(

Flags: needinfo?(mcastelluccio)

(In reply to Marco Castelluccio [:marco] from comment #5)

(In reply to Dave House [:dhouse] from comment #4)

(In reply to Marco Castelluccio [:marco] from comment #3)

Could you create the mapping for all commits? Or would that be expensive?
I'm doing some analyses on the history, so I'd need many of the old commits to be mapped.

I think we can create the mappings for all commits. There are a few waypoints where the conversion changed, and so it will take some time to work through the history.

When do you need the mappings for all commits? I will start working on the mapping between things, but unless you need it sooner I think it could be about a week (as I fit in time working on it between other tasks).

Ideally as soon as you can, as it is blocking me unfortunately :(

I started reading through the part of the vcssync code for the mapping, and the hggit extension. From what I have found, it creates the mapping when the commits are converted, and there does not appear to be a way to make it create the mapping from existing converted commits.

It may be simpler to take the two repos and write something to walk through the two commit-by-commit to create the mapping.

Hi Hal, re: vcssync do you know (or have an inkling) if there is a good way to create the hg-git mapping for old commits (I'm looking at pre-2005)?

It seemed like it should be "easy" but then I started reading through the code and I'm concluding that I need to read and match commit-by-commit myself.

Flags: needinfo?(hwine)

Marco, what I'm looking at is to write a script to walk through commits in gecko-dev and search for matching (commit time, and if that misses some then I'll look at author or fuzzy comment matching) commits in mozilla-central. Is that something that you would be better suited to write than me?

Flags: needinfo?(mcastelluccio)

(In reply to Dave House [:dhouse] from comment #8)

Marco, what I'm looking at is to write a script to walk through commits in gecko-dev and search for matching (commit time, and if that misses some then I'll look at author or fuzzy comment matching) commits in mozilla-central. Is that something that you would be better suited to write than me?

I can probably do that.
Wouldn't it be simpler to assume the same ordering of commits in the two repos? Or is that not the case?

Actually, pre-2007 we shouldn't have any hg commit, as there was CVS back then (and the CVS history is in the git repo but not in the mercurial one).

So, I'll close this as WONTFIX and will create a separate bug for the missing git commits that are after the beginning of the hg history.

Status: NEW → RESOLVED
Closed: 6 years ago
Flags: needinfo?(mcastelluccio)
Flags: needinfo?(hwine)
Resolution: --- → WONTFIX
See Also: → 1555175

(In reply to Marco Castelluccio [:marco] from comment #11)

So, I'll close this as WONTFIX and will create a separate bug for the missing git commits that are after the beginning of the hg history.

Thank you! That makes sense.

(In reply to Marco Castelluccio [:marco] from comment #10)

Actually, pre-2007 we shouldn't have any hg commit, as there was CVS back then (and the CVS history is in the git repo but not in the mercurial one).

Correct. If it would help, I believe there are some archived "initial CVS->git" conversion files. (As Dave noted, the conversion is lossy - there was no clean CVS->hg cutover.) There is also an archive of the CVS repo at time of decommission. And somewhere I have notes on the CVS->git conversion -- there were a lot of tradeoffs. Ping me if you need these, and I'll do a more thorough search.

Others who might have relevant CVS->git conversion lore include Aki & Ehsan. Good luck!

See Also: → 1568507
See Also: → 1569277
See Also: → 1595816
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: