Bug 5997

Summary: Build must work even if git is blocked by a firewall
Product: [QA/Testing] Test Plans/Suite Reporter: Dave Stewart <david.c.stewart>
Component: test-plans/suiteAssignee: Mihail Stanciu <stanciux.mihail>
Status: CLOSED FIXED QA Contact:
Severity: enhancement    
Priority: Medium+ CC: bluelightning, corneliux.stoicescu, poky.bs.watcher, poky.watcher, richard.purdie
Version: 1.6   
Target Milestone: 1.7 M4   
Hardware: All   
OS: Multiple   
Whiteboard:
OS type for building Yocto: --- Type of Regression: ---
Verified: Documentation change: No (bug/feature does not impact docs)

Description Dave Stewart 2014-03-18 14:55:20 UTC
If git is blocked due to proxy issues, the system should fall back to a source mirror. (In fact, I thought this worked in the past, so I believe this is a regression).

Make sure this is a permanent part of automated testing. If that's hard, add it to manual testing until it's automated.
Comment 1 Richard Purdie 2014-04-04 15:31:17 UTC
This has been confirmed to work on 1.6, marking this as an enhancement for the 1.7 testing plans.
Comment 2 Alexandru Georgescu 2014-07-30 08:29:09 UTC
after a short talk with Paul, a first test would be to block before a build the git using the firewall, then see if a build (e.g. minimal) can be completed using the fallback source mirrors.
Comment 3 Mihail Stanciu 2014-08-29 09:36:40 UTC
Sorry for this, but I am having trouble understanding the original intent of this test (or possibly something was lost in translation)

Do you want this test to check that:

1) if git isn't working, build completes by getting the needed packages by another method (wget, ftp or whatnot)

or

2) if the git source location (git.yoctoproject.org for example) isn't working, it switches to a different download mirror (say a local mirror you've set up in advance)

I hope I was coherent enough in my description.
Comment 4 Richard Purdie 2014-09-01 13:05:54 UTC
The first thing - that if git isn't working, the http mirrors have the necessary artifacts
Comment 5 Mihail Stanciu 2014-09-10 12:13:20 UTC
Test case proposed:
https://bugzilla.yoctoproject.org/tr_show_case.cgi?case_id=1058

Feedback would be appreciated.
Comment 6 Paul Eggleton 2014-09-18 15:14:40 UTC
Just looking over the test case:

1) It doesn't specify the starting condition - unless you already assume this for all test cases, you need to ensure you start with these recipes unfetched (so perhaps a bitbake -c cleanall <recipe> first, or alternatively state that you should start completely clean with an empty DL_DIR and TMPDIR)

2) Rather than offering a choice we should be explicit about what needs to be tested. If we're starting completely clean then I'd say bitbake -c fetchall core-image-sato would be a good target.
Comment 7 Mihail Stanciu 2014-09-22 12:05:27 UTC
Comments noted. Test case has been updated.

Marking as resolved.
Comment 8 Mihail Stanciu 2014-09-22 14:53:46 UTC
Closing.