Bug 11551

Summary: inconsistent dependencies in glibc-locale packages
Product: [Build System, Metadata & Runtime] OE-Core Reporter: Juro Bystricky <juro.bystricky>
Component: coreAssignee: Ross Burton <ross.burton>
Status: RESOLVED FIXED QA Contact:
Severity: normal    
Priority: Medium+ CC: meta.mr.watcher, meta.watcher, randy.macleod
Version: 2.3   
Target Milestone: 4.99   
Hardware: x86   
OS: Multiple   
Whiteboard:
OS type for building Yocto: --- Type of Regression: ---
Verified: Documentation change: No (bug/feature does not impact docs)
Attachments:
Description Flags
diffoscope output of two debian packages
none
diffoscope output of two RPM packages
none
diffoscope output comparing two builds. none

Description Juro Bystricky 2017-05-19 20:55:27 UTC
Created attachment 3767 [details]
diffoscope output of two debian packages

I built core-image-sato in two different folders, at the same time.
Same machine, settings, etc.
Yet I got a strange differences when comparing glibc-gconv-* packages.
(diffoscope files attached)
Debian packages show this:

build1: Depends:​·​glibc-​gconv
build2: Depends:​·​glibc-​gconv,​·​libc6·​(>=·​2.​25)​

RPM packages also show even more differences in REQUIRES fields.

This affected all glibc-gconv* packages.
After this I re-ran the build the third time. The third time the dependencies were the same as the second time (i.e. the "longer version")
Comment 1 Juro Bystricky 2017-05-19 20:56:29 UTC
Created attachment 3768 [details]
diffoscope output of two RPM packages
Comment 2 Ross Burton 2017-05-19 20:59:14 UTC
Really not shocked by this and I suspect its related to the permissions thing.
Comment 3 Ross Burton 2017-06-05 20:14:56 UTC
Further thoughts: I've seen glibc-locale not generate entire chunks of the packages before.
Comment 4 Juro Bystricky 2017-06-06 00:11:46 UTC
(In reply to comment #3)
> Further thoughts: I've seen glibc-locale not generate entire chunks of the
> packages before.

That is very interesting. Any bug report ?
Comment 5 Juro Bystricky 2017-09-15 22:10:45 UTC
The issue is still present, even after the commit
http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?h=master-next&id=06669f0f9e0f954ce08a958abde0639a4fa006aa
which I had high hopes for.
The buildhistories for the same package:

RDEPENDS = glibc (>= 2.26) glibc-gconv
RDEPENDS = glibc-gconv

Unfortunately, this affects all gconv packages (all good or bad at the same time).
Comment 6 Juro Bystricky 2018-02-06 22:36:48 UTC
Created attachment 4204 [details]
diffoscope output comparing two builds.
Comment 7 Juro Bystricky 2018-02-06 22:37:17 UTC
Still encountered on occasions even with 2.5 as of 385944254d0ef88bd6450a221a54cbcb40020b42

Depends:​·​glibc-​gconv,​·​libc6·​(>=·​2.​26)​
Depends:​·​glibc-​gconv

(see the attachement, two ipk packages)
Comment 8 Ross Burton 2019-09-16 20:22:52 UTC
I'm not convinced, but there's a chance this is related to the long standing glibc-locale determinism problem.  Closing unless we see it again.