Re: [Buildroot] Analysis of build results for 2021-08-17

From: Thomas Petazzoni <thomas.petazzoni@bootlin.com>
To: Giulio Benetti <giulio.benetti@benettiengineering.com>
Cc: Bernd Kuhls <bernd.kuhls@t-online.de>,
	Eric Le Bihan <eric.le.bihan.dev@free.fr>,
	Herve Codina <herve.codina@bootlin.com>,
	Matthew Weber <matthew.weber@rockwellcollins.com>,
	Gwenhael Goavec-Merou <gwenhael.goavec-merou@trabucayre.com>,
	Fabrice Fontaine <fontaine.fabrice@gmail.com>,
	buildroot@buildroot.org,
	Giulio Benetti <giulio.benetti@micronovasrl.com>,
	Adam Duskett <aduskett@gmail.com>
Subject: Re: [Buildroot] Analysis of build results for 2021-08-17
Date: Thu, 19 Aug 2021 15:28:52 +0200	[thread overview]
Message-ID: <20210819152852.760ec08d@windsurf> (raw)
In-Reply-To: <412e2832-62e5-8cbe-6a85-e797b3281978@benettiengineering.com>

Hello,

On Thu, 19 Aug 2021 00:24:06 +0200
Giulio Benetti <giulio.benetti@benettiengineering.com> wrote:

> This waits for you to release new OpenRisc Bootlin toolchain. At 
> Buildroot level all OpenRisc patches we're covered(on gcc 11.1.0 too 
> from yesterday). So with that you should be able to create the new 
> Bootlin OpenRisc Toolchain that will also enable -mcmodel=large for 
> libgeos and protobuf, basically they will be both re-enabled for 
> OpenRisc :-)

ACK, thanks for confirming.

> >>      or1k     |            unknown             | NOK | http://autobuild.buildroot.net/results/92c086b0c5c2a298a82a6eb9c289dc53c58a5e5e |  
> > 
> > Top-level parallel build, no idea what the issue is.  
> 
> Can somebody point me some idea on how to approach Top-level parallel 
> build debugging? I also have some udisks package failure due to that.
> I mean, now what I do is ./br_reproduce_build SHA1, then I CTRL-C and:
> # cd SHA1
> # cd output
> # make package-that-fails
> 
> But with Top-level I have more output-x, how can I reissue the command 
> and are there any trick to debug it?

I would go in two steps:

 (1) First see if you can reproduce without top-level parallel build,
     i.e you keep BR2_PER_PACKAGE_DIRECTORIES=y, but you do the build with
     "make". I am pretty sure that 95% of the issues will also be
     reproducible this way, and are more related to per-package
     directories than top-level parallel build.

 (2) If that doesn't allow to reproduce the issue, then you'll want to
     rebuild with "make -jX". However, here it would mean the issue is
     really with parallel build, so it's not because one build succeeds
     that another build won't fail: we might have race conditions that only
     rarely happen, just like the parallel build of individual packages. Of
     course, in principle with per-package directories, we are not supposed
     to have this kind of race conditions, but who knows what unforeseen
     issues still exist?

Best regards,

Thomas
-- 
Thomas Petazzoni, co-owner and CEO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
_______________________________________________
buildroot mailing list
buildroot@busybox.net
http://lists.busybox.net/mailman/listinfo/buildroot