Hello Albert, On Sat, 16 Jan 2016 12:21:06 +0100, Albert ARIBAUD <albert.u.b...@aribaud.net> wrote: > Hello Martin, > > On Fri, 15 Jan 2016 15:30:47 -0800, Martin Michlmayr <t...@cyrius.com> > wrote: > > * Albert ARIBAUD <albert.u.b...@aribaud.net> [2016-01-15 22:52]: > > > > Would you be so kind and test > > > > https://d-i.debian.org/daily-images/armel/daily/kirkwood/u-boot/openrd-client/ > > > > > > > > This is 2016.01 (no rc) > > > > > > It is the same as the one I tested first and which failed. I had > > > extracted it from > > > http://ftp.de.debian.org/debian/pool/main/u/u-boot/u-boot_2016.01+dfsg1-1_armel.deb > > > > Yes, they are the same. The d-i build process simply extracts the > > u-boot binary from the .deb, so it's easier for people to download. > > > > I'm surprised it fails to boot. Is 2016.01 from upstream working for > > you? > > I tested on the one hand the binary u-boot.kwb from d-i, and on the > other hand the binary built from the mainline U-Boot repo, tag > v2016-01, built with the toolchain fetched by U-Boot's buildman for > arm, which identifies itself as 'gcc version 4.9.0 (GCC)', using > binutils 2.24. The binary u-boot.kwb consistently fails to boot beyond > the few lines I gave. The buildman-built kwb runs consistently. > > > You can find the build log here, btw: > > https://buildd.debian.org/status/fetch.php?pkg=u-boot&arch=armel&ver=2016.01%2Bdfsg1-1&stamp=1452650663 > > Thanks. I see the build was done by newer gcc (5.3.1) and binutils > (2.25.90). I will set up a local buildd so that I can reproduce this( > and future builds an analyze them as needed).
Hmm, installing a complete buildd instance seems overkill. Is there a simple way to set up a build environment identical to that of the buildd which created u-boot.kwb? I've got VMs with basic Jessie, Testing and Unstable installs. Amicalement, -- Albert.