Repo clone tasks may fail due to 'svnsync' failing (remote server timeout or random error, etc). We should be smart about retrying those. Other failures that are a result of our code can happen too (failure to parse author, or timestamp properly) and it doesn't make sense to retry those.
I don't know about git/hg transfer errors. SVN seems to be the most common
On branch db/7857
To test, import an SVN repo and force an error (e.g. tweak command to use invalid arguments; or kill the svnsync command). Make sure that it logs appropriately and tries again. It should succeed if possible, or if it can't succeed then the failure email should still have the failure details.
I think we need to add description of new options (
scm.import.retry_count
&scm.import.retry_sleep_secs
) todevelopment.ini
.Email to
sfengineers@***.com
contains failure details, but shows "None" instead of import source (I guess this is old bug, but can be fixed now). Admin email shows source just fine.sfengineers email:
admin email:
Other than that all works fine for me.
Last edit: Dave Brondsema 2015-04-24
Added to development.ini in db/7857
Email contents come from an event handler hook, so fix for that is internal to SF in forge-classic:db/7857