On 5/1/25 05:13, Chris Hofstaedtler wrote:
On Wed, Apr 30, 2025 at 05:40:34PM -0400, Chris Knadle wrote:
This bug is a duplicate of #1103308

I've been working with upstream trying to figure out the test failures.

The code compiles fine but the tests time out on all but 4 architectures (2
release architectures).
It's 7 release architectures.
My statement above says the same thing, just differently. "All but 2" (i.e. all except 2 [out of 9]) is "7".
Unfortunately the test log seems to not have enough detail to debug
anything. However on arm64 at least I can see this happening:

bash -e ./tests/functional-tests discard
...
+ for d in incoming queue queue/discarded archive text subconf unsubconf bounce 
control moderation subscribers.d digesters.d requeue nomailsubs.d
+ mkdir list/nomailsubs.d
+ echo list@test
++ pwd
+ here=/home/ch/Debian/nmu/mlmmj-1.5.0
+ echo 'step 1'
step 1
+ atf_check -s exit:0 .../mlmmj-1.5.0/src/mlmmj-receive -L list -F
+ /usr/libexec/atf-check -s exit:0 
/home/ch/Debian/nmu/mlmmj-1.5.0/src/mlmmj-receive -L list -F
Executing command [ .../mlmmj-1.5.0/src/mlmmj-receive -L list -F ]

strace'ing mlmmj-receive shows a lot of this happening:

read(7, "", 4096)                       = 0
write(5, "\377", 1)                     = 1
read(7, "", 4096)                       = 0
write(5, "\377", 1)                     = 1
read(7, "", 4096)                       = 0
write(5, "\377", 1)                     = 1
read(7, "", 4096)                       = 0
write(5, "\377", 1)                     = 1
read(7, "", 4096)                       = 0
write(5, "\377", 1)                     = 1
read(7, "", 4096)                       = 0
write(5, "\377", 1)                     = 1
read(7, "", 4096)                       = 0
write(5, "\377", 1)                     = 1
read(7, "", 4096)                       = 0
write(5, "\377", 1)                     = 1

and mlmmj-receive taking one full CPU core until it gets killed.

The involved FDs are:

5u   mlmmj-1.5.0/list/queue/46edf6467742b084
6r   mlmmj-1.5.0/list/incoming/1d47fc2375333420
7r   mlmmj-1.5.0/list/incoming/1d47fc2375333420

% hexdump -C mlmmj-1.5.0/list/incoming/1d47fc2375333420
00000000  62 6c 61 0a                                       |bla.|
00000004

This seems like a code bug.

I'm attaching kyua logs and DB file from a build I did on an emulated arm64 cowbuilder buildd, in case those logs yield further useful information.

   -- Chris

Chris Knadle
chris.kna...@coredump.us

Attachment: kyua-mlmmj-1.5.0-arm64.tar.xz
Description: application/xz

Reply via email to