583
|
|
|
Nick Papior |
7 years ago
|
|
|
582
|
|
|
Nick Papior |
7 years ago
|
|
|
581
|
|
|
Nick Papior |
7 years ago
|
|
|
580
|
|
Merge 4.1-605, patchlogs only (with criss-cross)
This merge involved a criss-cross. From the bzr documentation:
Criss-crosses occur in a branch's history if two branches merge the same thing and then merge one another, or if two branches merge one another at the same time.
Criss-crosses cause problems because of the way merge works. Bazaar's default merge is a three-way merger; in order to merge OTHER into THIS, it must find a basis for comparison, BASE. Using BASE, it can determine whether differences between THIS and OTHER are due to one side adding lines, or from another side removing lines.
Criss-crosses mean there is no good choice for a base. Selecting the recent merge points could cause one side's changes to be silently discarded. Selecting older merge points (which Bazaar does) mean that extra conflicts are emitted.
The ``weave`` merge type is not affected by this problem because it uses line-origin detection instead of a basis revision to determine the cause of differences.
Indeed, when requesting the most recent common revision before the merge:
$ bzr log -rancestor:../4.1 --line 483.3.15: Alberto Garcia 2016-09-13 [merge] Changes to further standards compliance
the answer was quite an old one.
This is because the following patchlogs were already in the trunk, having been merged directly from 4.0 (at revisions 571 and 573):
------------------------ revno: 483.3.16 revision-id: nickpapior@gmail.com-20161006072528-03uhewjjhqwvc0s3 parent: albertog@icmab.es-20160913120042-79waws5ec5d2cc49 fixes bug: https://launchpad.net/bugs/1630827 committer: Nick Papior <nickpapior@gmail.com> branch nick: 4.0 timestamp: Thu 2016-10-06 09:25:28 +0200 message: Did not allow user control of SlabDipoleCorrection, fixes lp:1630827 ------------------------ revno: 483.3.17 revision-id: albertog@icmab.es-20161013124905-i8mcs29ht08lvsie parent: nickpapior@gmail.com-20161006072528-03uhewjjhqwvc0s3 fixes bug: https://launchpad.net/bugs/1625725 committer: Alberto Garcia <albertog@icmab.es> branch nick: 4.0 timestamp: Thu 2016-10-13 14:49:05 +0200 message: Fix bug 1625725 in 'nodes' basis-generation option ------------------------
They were also included in the to-be-merged revno 605 in 4.1, which had itself merged from 4.0 followed by "revert ." to sync the patchlogs.
The "common ancestor" was reported as 483.3.15 since this is:
revno: 483.3.15 [merge] revision-id: albertog@icmab.es-20160913120042-79waws5ec5d2cc49
the parent of 483.3.16 (see above), the oldest of the patchlogs to be merged.
With hindsight, the direct merge in the trunk from 4.0 (with 4.1 already active and needing the same fixes) was ill-advised. Here is an edited summary of ideas exchanged while trying to establish a future policy about this:
----------------------------------------------------------------- Assuming that release branches are only for bug fixes:
If a bug is discovered in 4.0 which does not apply to 4.1, it does not make sense to merge *the substance* of it into 4.1. It is a question of cosmetics whether *the patchlogs* should be merged, so that there is a clear record that all bugs discovered in past release branches have been taken care of in newer branches (either applied in substance or marked as "does not apply" by just importing the patchlogs).
On the other hand, if the bug does apply, I think that the "commit only once in the repository" philosophy would require that the bug is committed first in 4.0, and (the same patch) merged into 4.1. For example, if you do now, in 4.1 (this was at revno 604):
bzr missing --line ../4.0
you will get:
[...] You are missing 2 revisions: 503: Alberto Garcia 2016-10-13 Fix bug 1625725 in 'nodes' basis-generation option 502: Nick Papior 2016-10-06 Did not allow user control of SlabDipoleCorrection, fixes ...
but 4.1 does have the *substance* of the fix . It seems that the file changes from 4.0 were replicated in 4.1 without merging the patchlogs. Now one needs to double-check that the fix is actually applied in 4.1, either by "bzr log" (hoping that the committers are consistent in the subject lines) or in the code...
There might be problems with transportation of unwanted changes (bits that do not quite apply), but this should be taken care of at the time of merging into 4.1. The net result is that there will be a single "main substance" commit for each "bug report" that applies to both versions. This commit will go into 4.0, and be merged (with the appropriate modifications) into 4.1, and then on into trunk (again, with the appropriate modifications, as always). -------------------------------------------------------end
NOTE: As an example of a patchlog that "does not quite apply", here is a fix in 4.0 for an issue that had already been corrected in 4.1 (see the bug page for more details):
-------------------- revno: 483.3.18 fixes bug: https://launchpad.net/bugs/1636100 committer: Nick Papior <nickpapior@gmail.com> branch nick: 4.0 timestamp: Mon 2016-10-24 09:27:38 +0200 message: Fixes non-colinear band calculation, fixes lp:1636100 --------------------
This patchlog comes to trunk via the merge of 4.0 patchlogs into 4.1.
This merge was carried out with the command
bzr merge --weave ../4.1
which did minimize the conflicts involved. It should be noted that the offending repeated patchlogs are not incorporated again into the branch history. (You can check this by doing bzr log -r 580 -n0 )
Hopefully future merges will not cause problems.
|
Alberto Garcia |
7 years ago
|
|
|
579
|
|
|
Alberto Garcia |
7 years ago
|
|
|
578
|
|
|
Alberto Garcia |
7 years ago
|
|
|
577
|
|
|
Alberto Garcia |
7 years ago
|
|
|
576
|
|
|
Nick Papior |
7 years ago
|
|
|
575
|
|
|
Nick Papior |
7 years ago
|
|
|
574
|
|
|
Nick Papior |
7 years ago
|
|
|
573
|
|
|
Nick Papior |
7 years ago
|
|
|
572
|
|
|
Nick Papior |
7 years ago
|
|
|
571
|
|
|
Nick Papior |
7 years ago
|
|
|
570
|
|
|
Nick Papior |
7 years ago
|
|
|
569
|
|
|
Rafi Ullah |
7 years ago
|
|
|
568
|
|
|
Alberto Garcia |
7 years ago
|
|
|
567
|
|
|
Alberto Garcia |
7 years ago
|
|
|
566
|
|
|
Alberto Garcia |
7 years ago
|
|
|
565
|
|
|
Alberto Garcia |
7 years ago
|
|
|
564
|
|
|
Alberto Garcia |
7 years ago
|
|
|