| Summary: | Use "yocto-kernel patch add" added a patch, then core-image-sato build hangs. | ||||||
|---|---|---|---|---|---|---|---|
| Product: | [Build System, Metadata & Runtime] Meta-yocto | Reporter: | Ning Zhang <ning.zhang> | ||||
| Component: | meta-yocto | Assignee: | Bruce Ashfield <bruce.ashfield> | ||||
| Status: | VERIFIED FIXED | QA Contact: | |||||
| Severity: | normal | ||||||
| Priority: | Medium+ | CC: | alexandrux.palalau, bruce.ashfield, ke.zou, poky.bs.watcher, poky.watcher, sgw, tom.zanussi, yi.zhao | ||||
| Version: | 1.5 | ||||||
| Target Milestone: | 1.5 M5 | ||||||
| Hardware: | All | ||||||
| OS: | Multiple | ||||||
| Whiteboard: | |||||||
| OS type for building Yocto: | --- | Type of Regression: | --- | ||||
| Verified: | Documentation change: | Don't know | |||||
| Attachments: |
|
||||||
|
Description
Ning Zhang
2013-09-03 07:07:11 UTC
It looks like this might be a regression in the kernel tools. From linux-yocto log.do_patch, this goes on forever: No author found in No author found in No author found in No author found in No author found in No author found in . . . Possibly an infinite loop in git-s2q introduced by the following commit? commit f9e85358bd9d6e583253523c9182268f64bafc27 Author: Paul Gortmaker <paul.gortmaker@windriver.com> Date: Thu Aug 22 15:14:57 2013 -0400 kgit-s2q: continue if git-am returns zero Let's check with Bruce and see if this could be the problem before digging further.. The tools have changed a lot .. badly formatted patches are always a challenge. I'll take care of testing this out and coming up with a fix. We don't want to regress the rest of the cases, fixing this! I think the patch is good, I'm not sure what could be badly formatted about it, but anyway yeah, it's never caused a blip before.. Some comments: 1) My commit f9e85358bd9d6e was a code relocation block, and not a functional change, so it is irrelevant here. 2) The "No author found in <missing text>" is because there is a left over, unassigned "patch_name" variable that has been in existence since this code originally borrowed the same block of code from git quilt-import. 3) The block of code above is only ever entered when the system is fed a garbage commit that has no date, no author and/or no subject, in an attempt to get that info interactively. 4) We'll want to delete that possibility of an interactive step, since it doesn't bode well with automated builds. 5) Finally, the patch posted here is just that -- a patch and not a commit. Hence it does not have the required fields of date/author/subject and is destined to fail since it is invalid input. We'll want to fix (i.e. remove) the possibility of going into an interactive phase, but that does not change the fact the patch in question here is invalid. I didn't blame that commit, just a possibility pointed out by git blame. I didn't realize that all patches in Yocto needed to have Author, Date and Subject in order to apply. I'll have to verify, but I'm pretty sure this test case passed before. QA should be able confirm or not. If not, they should be using a 'valid patch'. I have a fix, will send the pull request shortly. This is fixed, but I'm waiting for the rc build to be done, before submitting the fix, since it'll just wait. Verified it on master branch. Commit ID: e7a20238dc73bb449f15feaf0dead51daf796184 3.10 kernel, though |