-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of
Martyn Welch
Sent: Wednesday, January 26, 2011 4:33 AM
To: [email protected]
Subject: Re: [oe] bitbake does not fail when QA issues encountered
On 25/01/11 21:36, Khem Raj wrote:
On Tue, Jan 25, 2011 at 12:58 PM, Frans Meulenbroeks
<[email protected]> wrote:
2011/1/25 Maupin, Chase<[email protected]>:
All,
I have noticed that when building packages such as perl that while my
build will report success and no errors, the return status from the
bitbake command was "1". I was able to produce this by doing:
MACHINE=am37x-evm bitbake perl
After bitbake completed I saw:
NOTE: Tasks Summary: Attempted 851 tasks of which 0 didn't need to be
rerun and 0 failed.
but checking $? yields a return status of "1".
I looked into the log and noticed a lot of messages like:
ERROR: QA Issue with db: package db contains bad RPATH
My understanding is that recent fixes to libtool 2.4 prevent these
errors but I am using an older version of Angstrom which pins to libtool
2.2. I also have found this issue with the Arago distribution which
likewise uses libtool 2.2.
So my question here is whether bitbake should be failing when it
encounters these QA issues with a bad RPATH and exiting?
If not then should the return status be "1"? This causes issues when
using a script that issues builds and then checks the return status for
success or failure. If the QA issues are deemed acceptable (or should be
warnings) then I would expect the return status to not indicate a failure.
I have attached a log of my build for reference
As another interesting side note which I don't know is related or not,
when building Arago with bitbake 1.10.2 the return status is "1". When
building the same Arago metadata with bitbake 1.8.19 the return status is
"0". What is strange here is that since Arago uses a slightly older
version of the OE metadata it is not seeing the RPATH errors reported
above (the check isn't in the insane.bbclass for Arago yet). So for some
reason bitbake 1.8.19 says everything went fine and bitbake 1.10.2 reports
a status of "1" even though there is no reported error. I'm not sure if
this is related to the above in any way or if this is a separate issue.
Sincerely,
Chase Maupin
I've seen this on other places as well.
I'd say if a package has a QA issue the build of that package should
fail, because the resulting output is defnitely not OK.
yes it should fail. However some may raise questions "it used to build
and not it doesnt"
so someone has to fix the problems quickly
...and if it is considered a failure and returns 1, the summary shouldn't
be
reporting "0 failed", or at least there should be something reported at
the
end of the build to state that the build has been deemed a failure for
those
not running in a script and who don't read through the entire log of the
build!
