imx8ulp_evk build with buildman fails every other time

[Simon Glass]

On Wed, 28 Aug 2024 at 09:40, Jerome Forissier
<jerome.forissier at linaro.org> wrote:
>
>
>
> On 8/27/24 17:53, Simon Glass wrote:
> > Hi Jerome,
> >
> > On Tue, 27 Aug 2024 at 08:53, Fabio Estevam <festevam at gmail.com> wrote:
> >>
> >> Adding Gary and Peng on Cc.
> >>
> >> On Tue, Aug 27, 2024 at 11:42 AM Jerome Forissier
> >> <jerome.forissier at linaro.org> wrote:
> >>>
> >>> Hi,
> >>>
> >>> Can someone explain what's happening with the imx8ulp_evk buildman build
> >>> below? It either fails with a mkimage error, or succeeds with a binman
> >>> warning, every other time. 100% reproducible. /me confused :o
> >>>
> >>> ====8<============8<==============
> >>> $ docker run -it --rm trini/u-boot-gitlab-ci-runner:jammy-20240808-21Aug2024
> >>> uboot at 42aa85c2d6b7:/$ cd
> >>> uboot at 42aa85c2d6b7:~$ git clone https://github.com/u-boot/u-boot.git
> >>> uboot at 42aa85c2d6b7:~/u-boot$ git log --oneline -n 1
> >>> ee2af844ba (HEAD -> master, origin/master, origin/HEAD) Merge tag 'efi-2024-10-rc4' of https://source.denx.de/u-boot/custodians/u-boot-efi
> >>> uboot at 42aa85c2d6b7:~$ cd u-boot
> >>> uboot at 42aa85c2d6b7:~/u-boot$ tools/buildman/buildman imx8ulp_evk
> >>> Building current source for 1 boards (1 thread, 24 jobs per thread)
> >>>    aarch64:  w+   imx8ulp_evk
> >>> +WARNING 'mx8ulpa2-ahab-container.img' not found, resulting binary may be not-functional
> >>> +WARNING 'bl31.bin' not found, resulting binary may be not-functional
> >>> +WARNING 'upower.bin' not found, resulting binary may be not-functional
> >>> +WARNING 'm33_image.bin' not found, resulting binary may be not-functional
> >>>     0    1    0 /1              imx8ulp_evk
> >>> Completed: 1 total built, 1 newly), duration 0:00:09, rate 0.11
> >>> uboot at 42aa85c2d6b7:~/u-boot$ tools/buildman/buildman imx8ulp_evk
> >>> Building current source for 1 boards (1 thread, 24 jobs per thread)
> >>>    aarch64:  +   imx8ulp_evk
> >>> +WARNING 'bl31.bin' not found, resulting binary may be not-functional
> >>> +binman: Error 1 running 'mkimage -d ./mkimage.spl.mkimage -n spl/u-boot-spl.cfgout -T imx8image -e 0x22020000 ./mkimage-out.spl.mkimage': Fail open first container file mx8ulpa2-ahab-container.img
> >>> +
> >>> +make[1]: *** [Makefile:1126: .binman_stamp] Error 1
> >>> +make: *** [Makefile:177: sub-make] Error 2
> >>>     0    0    1 /1              imx8ulp_evk
> >>> Completed: 1 total built, 1 newly), duration 0:00:04, rate 0.25
> >>> uboot at 42aa85c2d6b7:~/u-boot$ tools/buildman/buildman imx8ulp_evk
> >>> Building current source for 1 boards (1 thread, 24 jobs per thread)
> >>>    aarch64:  w+   imx8ulp_evk
> >>> +WARNING 'mx8ulpa2-ahab-container.img' not found, resulting binary may be not-functional
> >>> +WARNING 'upower.bin' not found, resulting binary may be not-functional
> >>> +WARNING 'm33_image.bin' not found, resulting binary may be not-functional
> >>> +WARNING 'bl31.bin' not found, resulting binary may be not-functional
> >>>     0    1    0 /1              imx8ulp_evk
> >>> Completed: 1 total built, 1 newly), duration 0:00:04, rate 0.25
> >>> uboot at 42aa85c2d6b7:~/u-boot$ tools/buildman/buildman imx8ulp_evk
> >>> Building current source for 1 boards (1 thread, 24 jobs per thread)
> >>>    aarch64:  +   imx8ulp_evk
> >>> +WARNING 'bl31.bin' not found, resulting binary may be not-functional
> >>> +binman: Error 1 running 'mkimage -d ./mkimage.spl.mkimage -n spl/u-boot-spl.cfgout -T imx8image -e 0x22020000 ./mkimage-out.spl.mkimage': Fail open first container file mx8ulpa2-ahab-container.img
> >>> +
> >>> +make[1]: *** [Makefile:1126: .binman_stamp] Error 1
> >>> +make: *** [Makefile:177: sub-make] Error 2
> >>>     0    0    1 /1              imx8ulp_evk
> >>> Completed: 1 total built, 1 newly), duration 0:00:04, rate 0.25
> >>> ====8<============8<==============
> >>>
> >
> > I can't quite explain it but I have hit this before. It is why I have
> > been unable to land [1]
> >
> > Perhaps binman is writing a file which is being picked up on the next
> > run, but is not valid? Or perhaps the file from a previous build
> > confuses things? I notice that some of the sections have filenames,
> > meaning that they are written and persist between builds.
> >
> > You can try BINMAN_VERBOSE=4 or 5 to see what is happening, or
> > BINMAN_DEBUG=1 to get a stack trace.
>
> That one wasn't super easy to track down :-/ but I think I have a proper
> fix. See https://lists.denx.de/pipermail/u-boot/2024-August/563355.html.

Nice work!!

- Simon

> >
> > [1] https://patchwork.ozlabs.org/project/uboot/patch/20240623175515.1466908-14-sjg@chromium.org/