publish_tarball is a sad bunny 🐰

35 views
Skip to first unread message

Evangelos Foutras

unread,
Feb 25, 2021, 1:48:31 AMFeb 25
to chromium-packagers
Lots of "download pnacl toolchain dependencies" failures in the past week. It also seems to have uploaded incomplete sets of the exported tarballs (e.g. 89.0.4389.58 which is being retried now).

Can someone look into this? (Hopefully the backlog won't be big enough to delay next's week stable release.)

Kubo Da Costa, Raphael

unread,
Feb 25, 2021, 5:51:28 AMFeb 25
to chromium-...@chromium.org, fabian...@chromium.org, dsc...@chromium.org
CC'ing some of the folks who seem to be involved in this area based on the latest commits I could find.

As far as I can see, the problem's that according to https://chromium.googlesource.com/a/native_client/nacl-llvm-project-v10/+/8b34291f6d0b5e887747cf7ecb7e14ec1cdafc35 there was a rebase that rewrote the commit history in that repository, and none of the previous hashes for "llvm_saigo" in https://chromium.googlesource.com/native_client/src/native_client/+/refs/heads/master/pnacl/COMPONENT_REVISIONS seem to be retrievable anymore.

PNaCl folks: is there a way to restore those old commits and preserve them in a branch, and to prevent future rebases from erasing old commits as well?

Kubo Da Costa, Raphael

unread,
Feb 26, 2021, 3:42:37 AMFeb 26
to dsc...@google.com, fabian...@chromium.org, chromium-...@chromium.org
I see that the publish_tarball bot seems to have gone green again, so thank you for fixing it so quickly!

On Thu, 2021-02-25 at 10:49 -0800, Derek Schuff wrote:
(replying again to the group):

Sorry about that. I agree that the commits shouldn't have just disappeared, and it looks like Fabian has already pushed them back up to chromium.googlesource.com on the 'saigo-v11' branch.

Note that the "llvm_saigo" compiler is not yet used for anything (in the Chrome build or otherwise), so it should be possible to build a release of chrome (including any source tarballs) without it. The plan is for it to eventually replace the current nacl-clang as the compiler that builds the NaCl IRT target.

Evangelos Foutras

unread,
Feb 26, 2021, 5:06:21 AMFeb 26
to Kubo Da Costa, Raphael, chromium-...@chromium.org
Awesome, thanks!

Another (unrelated) issue is that the bot_update step sometimes takes hours to complete (e.g. [1]; ~5 hours so far). It's usually fast (~5 minutes) but I've seen it being very slow in the past too.

It appears to be hitting a curl timeout while cloning src.git, due to low transfer speeds, aborting just as the remote end is getting ready to send data. It is treated as a transient error and the clone is retried, but it often fails several times before it finally succeeds.


--
You received this message because you are subscribed to the Google Groups "chromium-packagers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to chromium-packag...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/chromium-packagers/abb56e02f2ac452f603605cd419e5aa977390e41.camel%40intel.com.

Kubo Da Costa, Raphael

unread,
Feb 26, 2021, 5:24:51 AMFeb 26
to evan...@foutrelis.com, thomasa...@chromium.org, chromium-...@chromium.org
+thomasanderson, as I don't think there's much I can do about this one (or even if it's just The Cloud acting out :-)

Dirk Pranke

unread,
Feb 26, 2021, 3:34:10 PMFeb 26
to Tom Anderson, Kubo Da Costa, Raphael, evan...@foutrelis.com, thomasa...@chromium.org, chromium-...@chromium.org
I would really only expect that if we didn't have the caching set up properly, so file a bug w/ go/bug-a-trooper for someone to take a look.

-- Dirk

On Fri, Feb 26, 2021 at 12:13 PM Tom Anderson <thomasa...@google.com> wrote:
Looping in someone from infra.

+Dirk Pranke is it a known issue that bot_update sometimes takes a very long time (5 hours):


--
You received this message because you are subscribed to the Google Groups "chromium-packagers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to chromium-packag...@chromium.org.

Tom Anderson

unread,
Mar 8, 2021, 3:18:12 PMMar 8
to Kubo Da Costa, Raphael, Dirk Pranke, evan...@foutrelis.com, thomasa...@chromium.org, chromium-...@chromium.org
Looping in someone from infra.

+Dirk Pranke is it a known issue that bot_update sometimes takes a very long time (5 hours):


--
You received this message because you are subscribed to the Google Groups "chromium-packagers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to chromium-packag...@chromium.org.

Derek Schuff

unread,
Mar 8, 2021, 3:18:12 PMMar 8
to Kubo Da Costa, Raphael, chromium-...@chromium.org, fabian...@chromium.org
(replying again to the group):

Sorry about that. I agree that the commits shouldn't have just disappeared, and it looks like Fabian has already pushed them back up to chromium.googlesource.com on the 'saigo-v11' branch.

Note that the "llvm_saigo" compiler is not yet used for anything (in the Chrome build or otherwise), so it should be possible to build a release of chrome (including any source tarballs) without it. The plan is for it to eventually replace the current nacl-clang as the compiler that builds the NaCl IRT target.

On Thu, Feb 25, 2021 at 2:51 AM Kubo Da Costa, Raphael <raphael.ku...@intel.com> wrote:
Reply all
Reply to author
Forward
0 new messages