I've tried again with the change from Alex (without additional SRCREV bump)
on ubuntu-14.04

To make it simple I've triggered the build of just to recipes (which caused
"tried to close client 0 (highest is 1)" before:

$ bitbake tzdata update-rc.d
and again I quickly got:
23G     tzdata/2018c-r0/pseudo/
24G     update-rc.d/0.7-r5/pseudo/

There is nothing special in the pseudo.log before this "tried to close
client 0 (highest is 1)"
==> tzdata/2018c-r0/pseudo/pseudo.log <==
debug_logfile: fd 2
pid 8489 [parent 8484], doing new pid setup and server start
Setup complete, sending SIGUSR1 to pid 8484.
tried to close client 0 (highest is 1)
tried to close client 0 (highest is 1)

==> update-rc.d/0.7-r5/pseudo/pseudo.log <==
debug_logfile: fd 2
pid 8488 [parent 8481], doing new pid setup and server start
Setup complete, sending SIGUSR1 to pid 8481.
tried to close client 0 (highest is 1)
tried to close client 0 (highest is 1)

And there is actually nothing else after those 3 lines at the beginning
until I've killed the bitbake build (which didn't kill pseudo itself) and
then killed both pseudo processes - which resulted with left-over
bitbake-worker starting 2 new pseudo processes which kept spamming "tried
to close client 0 (highest is 1)" after initial setup:

$grep -v "^tried to close client 0 (highest is 1)$" */*/pseudo/pseudo.log
tzdata/2018c-r0/pseudo/pseudo.log:debug_logfile: fd 2
tzdata/2018c-r0/pseudo/pseudo.log:pid 8489 [parent 8484], doing new pid
setup and server start
tzdata/2018c-r0/pseudo/pseudo.log:Setup complete, sending SIGUSR1 to pid
8484.
tzdata/2018c-r0/pseudo/pseudo.log:Received signal 15, quitting.
tzdata/2018c-r0/pseudo/pseudo.log:db cleanup for server shutdown,
19:10:01.424
tzdata/2018c-r0/pseudo/pseudo.log:memory-to-file backup complete,
19:10:01.436.
tzdata/2018c-r0/pseudo/pseudo.log:db cleanup finished, 19:10:01.436
tzdata/2018c-r0/pseudo/pseudo.log:debug_logfile: fd 2
tzdata/2018c-r0/pseudo/pseudo.log:pid 11205 [parent 11199], doing new pid
setup and server start
tzdata/2018c-r0/pseudo/pseudo.log:Setup complete, sending SIGUSR1 to pid
11199.
update-rc.d/0.7-r5/pseudo/pseudo.log:debug_logfile: fd 2
update-rc.d/0.7-r5/pseudo/pseudo.log:pid 8488 [parent 8481], doing new pid
setup and server start
update-rc.d/0.7-r5/pseudo/pseudo.log:Setup complete, sending SIGUSR1 to pid
8481.
update-rc.d/0.7-r5/pseudo/pseudo.log:Received signal 15, quitting.
update-rc.d/0.7-r5/pseudo/pseudo.log:db cleanup for server shutdown,
19:10:01.423
update-rc.d/0.7-r5/pseudo/pseudo.log:memory-to-file backup complete,
19:10:01.436.
update-rc.d/0.7-r5/pseudo/pseudo.log:db cleanup finished, 19:10:01.436
update-rc.d/0.7-r5/pseudo/pseudo.log:debug_logfile: fd 2
update-rc.d/0.7-r5/pseudo/pseudo.log:pid 11212 [parent 11200], doing new
pid setup and server start
update-rc.d/0.7-r5/pseudo/pseudo.log:Setup complete, sending SIGUSR1 to pid
11200.

$ du -hs */*/pseudo/
28G     tzdata/2018c-r0/pseudo/
28G     update-rc.d/0.7-r5/pseudo/

Without --enable-epoll in pseudo.inc tzdata built fine and pseudo.log is
very short:
debug_logfile: fd 2
pid 14630 [parent 14629], doing new pid setup and server start
14630: new pid: 14630 [pseudo]
14630: Setup complete, sending SIGUSR1 to pid 14629.
14630: path mismatch [2 links]: ino 282238092 db '/run/shm/sem.mp-yluz7s8j'
req '/run/shm/sem.5mP74h'.
14630: db cleanup for server shutdown, 20:19:27.475
14630: memory-to-file backup complete, 20:19:27.475.
14630: db cleanup finished, 20:19:27.475

Shouldn't PSEUDO_DEBUG=4 in the environment generate a bit more verbose log?

I've also tried to add it in FAKEROOTENV (because pseudo seems to lock up
before run.do_install is executed, so the export there probably didn't
happen soon enough, but then even when it worked without --enable-epoll the
log wasn't any bigger).

meta/conf/bitbake.conf:export PSEUDO_DEBUG = "4"
meta/conf/bitbake.conf:FAKEROOTBASEENV =
"PSEUDO_BINDIR=${PSEUDO_SYSROOT}${bindir_native}
PSEUDO_LIBDIR=${PSEUDO_SYSROOT}${prefix_native}/lib/pseudo/lib
PSEUDO_PREFIX=${PSEUDO_SYSROOT}${prefix_native} PSEUDO_DISABLED=1
PSEUDO_DEBUG=4"
meta/conf/bitbake.conf:FAKEROOTENV =
"PSEUDO_PREFIX=${PSEUDO_SYSROOT}${prefix_native}
PSEUDO_LOCALSTATEDIR=${PSEUDO_LOCALSTATEDIR} PSEUDO_PASSWD=${PSEUDO_PASSWD}
PSEUDO_NOSYMLINKEXP=1 PSEUDO_DEBUG=4 PSEUDO_DISABLED=0"

Regards,

On Sun, Feb 18, 2018 at 12:43 PM, Martin Jansa <martin.jansa@gmail.com>
wrote:

> The do_install tasks failing with exit code '134'/SIGABRT were caused by
> left over pseudo processes from previous builds still holding the
> pseudo.lock.
>
> Manually killing those (still spinning at around 90% CPU for last 2 days)
> resolved the do_install failures.
>
> It also explains the lack of log.do_install files, because do_install
> failed before it even properly started.
>
> Regards,
>
> On Sat, Feb 17, 2018 at 9:17 PM, Seebs <seebs@seebs.net> wrote:
>
>> On Sat, 17 Feb 2018 15:28:55 +0200
>> Alexander Kanavin <alexander.kanavin@linux.intel.com> wrote:
>>
>> > For me, with epoll enabled:
>> >
>> > b6a015a works
>> > 23f089f and 26e30fa both lock up
>>
>> Huh. It's possible that the initial "don't try to close fd 0" was
>> correct, and the real problem is that the attempt is getting made
>> mistakenly. I'll study that more; the epoll code was a contribution and
>> I may not have fully understood it.
>>
>> -s
>>
>
>