git show shall compare with last build sha1

[blatinville]

use git show --no-abbrev --format=raw -M --raw 'old sha1'..'new sha1'
instead of
git show --no-abbrev --format=raw -M --raw 'new sha1'

This is the correction for
"Git plug-in & included regions : build not triggered in case of mutiple commits in one push."
https://issues.jenkins-ci.org/browse/JENKINS-13368.

I think think this correction also solves
"git plugin "included regions" feature is bypassed when evaluating a merge commit"
https://issues.jenkins-ci.org/browse/JENKINS-13580

We shall find a common correction for both issues are they are linked to include region feature.

[buildhive]

Jenkins » git-plugin #12 UNSTABLE
Looks like there's a problem with this pull request
(what's this?)

[blatinville]

Oops .. I could not checked the test yet. Test are not even running properly in my environment without any changes.

[blatinville]

I do not understand why so many test were failed. After clean build I get two failed test.

testBasicExcludedUser is failed "scm polling did not detect commit3 change" because the plugin
now list all changes since last build.
As there is one excluded user in the list of commits, the build is not run even though there
is one non-excluded user in the list. This is because the check on excluded user filers
exit as soon as one excluded user is found. It shall be changed to check if at least one
non excluded user is found.

testBasicExcludedUser is on error, did not had a look at it yet.

I had no time yet to spend more time on these issues. Looking at the code it become complex if
several branches/repositories are used. Not sure whether included/excluded region should be considered in case
of multiple branches/repositories.
On our Jenkins setup, we are using only one branch per job, using included regions (not excluded regions or users) and the correction has been working.

[cfoote]

See pull request #89 for another way to fix this.

[blatinville]

I'll check & test pull request #89 to see if it solves this problem.

[blatinville]

I've made basic testing pull request #89 and confirm that it also fixes the problem.
I'll test this in real conditions and close this pull request if I do not find any issues.

[blatinville]

I've tested Pull request #89 and I have the expected behaviour. As this pull request is older and not aligned with the latest git plugin code, I'll close it.