Pkg building on ampere* seems paused

[Ronald Klop]

Hi,

Looking at https://pkg-status.freebsd.org/builds?all=1&type=package and search 
for “ampere” it seems that no arm64 builds have been started since July.

Is this on purpose or is something broken?

Regards,
Ronald. 

bugzilla: add url link to a word

[Nuno Teixeira]

Hello all,

How can I use a [word] that links to a url in bugzilla?
e.g.:
278921 

Thanks,
-- 
Nuno Teixeira
FreeBSD UNIX: Web:  https://FreeBSD.org

Re: Pkg building on ampere* seems paused

[Philip Paeps]

On 2024-08-10 17:29:57 (+0800), Ronald Klop wrote:
Looking at https://pkg-status.freebsd.org/builds?all=1&type=package 
and search for “ampere” it seems that no arm64 builds have been 
started since July.


Is this on purpose or is something broken?


Sorry.  I broke something.  I paused builds for an upgrade in July, then 
got distracted fighting other cluster fires, and forgot to turn builds 
on again.  I've cleared the stop-builds flag on ampere* now.  I've also 
made a note to monitor this flag and yell when Philip gets distracted.


Apologies.

Philip

Re: Pkg building on ampere* seems paused

[Mark Millard]

Philip Paeps  wrote on
Date: Sat, 10 Aug 2024 15:06:33 UTC :

> On 2024-08-10 17:29:57 (+0800), Ronald Klop wrote:
> > Looking at https://pkg-status.freebsd.org/builds?all=1&type=package 
> > and search for “ampere” it seems that no arm64 builds have been 
> > started since July.
> >
> > Is this on purpose or is something broken?
> 
> Sorry. I broke something. I paused builds for an upgrade in July, then 
> got distracted fighting other cluster fires, and forgot to turn builds 
> on again. I've cleared the stop-builds flag on ampere* now. I've also 
> made a note to monitor this flag and yell when Philip gets distracted.

Has ampere2 had its jails updated so that main-armv7-default
has a chance of completing it build once it starts?

Note for Ronald K.: The need to update that jail's world
contributed to why ampere2 needed to be paused.
main-armv7-default looks to last have completed a "bulk -a"
with one which "ended at Sat Jun 1 08:10:26 UTC 2024".
After that, builds would hang up whenever the system loaded
a particular library (recursive use of a non-recursive lock
during the load). The "bulk -a" build would have to be
killed.

===
Mark Millard
marklmi at yahoo.com

Re: Pkg building on ampere* seems paused

[Mark Millard]

On Aug 10, 2024, at 09:21, Mark Millard  wrote:

> Philip Paeps  wrote on
> Date: Sat, 10 Aug 2024 15:06:33 UTC :
> 
>> On 2024-08-10 17:29:57 (+0800), Ronald Klop wrote:
>>> Looking at https://pkg-status.freebsd.org/builds?all=1&type=package 
>>> and search for “ampere” it seems that no arm64 builds have been 
>>> started since July.
>>> 
>>> Is this on purpose or is something broken?
>> 
>> Sorry. I broke something. I paused builds for an upgrade in July, then 
>> got distracted fighting other cluster fires, and forgot to turn builds 
>> on again. I've cleared the stop-builds flag on ampere* now. I've also 
>> made a note to monitor this flag and yell when Philip gets distracted.
> 
> Has ampere2 had its jails updated so that main-armv7-default
> has a chance of completing it build once it starts?
> 
> Note for Ronald K.: The need to update that jail's world
> contributed to why ampere2 needed to be paused.
> main-armv7-default looks to last have completed a "bulk -a"
> with one which "ended at Sat Jun 1 08:10:26 UTC 2024".
> After that, builds would hang up whenever the system loaded
> a particular library (recursive use of a non-recursive lock
> during the load). The "bulk -a" build would have to be
> killed.

I stupidly looked up the last failed "bulk -a" for
main-armv7-default instead of the last from-scratch
"bulk -a" that was successful (so the last known
guaranteed successful load of the library in
question):

pkg (first thing built) reported:

build started at Mon Feb 19 12:47:46 UTC 2024

and the last builder finished, reporting:

ended at Sat Feb 24 02:10:40 UTC 2024

So, over 5 months have gone by since then.

===
Mark Millard
marklmi at yahoo.com

perl build failure?

[Steve Kargl]

How is perl 5.36.3 the default version when it dies with

WARNING: re-sorting MANIFEST
*** [MANIFEST.srt] Error code 1

make: stopped making "all" in /usr/ports/lang/perl5.36/work/perl-5.36.3
--- lib/auto/re/re.so ---
Unsuccessful make(ext/re): code=1536 at make_ext.pl line 584.
*** [lib/auto/re/re.so] Error code 25

make: stopped making "all" in /usr/ports/lang/perl5.36/work/perl-5.36.3
make: 2 errors

make: stopped making "all" in /usr/ports/lang/perl5.36/work/perl-5.36.3
===> Compilation failed unexpectedly.
Try to set MAKE_JOBS_UNSAFE=yes and rebuild before reporting the failure to
the maintainer.
*** Error code 1

Stop.
make[1]: stopped making 
"/usr/ports/lang/perl5.36/work/.stage_done.perl._usr_local" in 
/usr/ports/lang/perl5.36
*** Error code 1

Stop.
make: stopped making "all" in /usr/ports/lang/perl5.36

-- 
Steve

Re: perl build failure?

[Steve Kargl]

On Sat, Aug 10, 2024 at 08:05:25PM -0700, Steve Kargl wrote:
> How is perl 5.36.3 the default version when it dies with
> 
> WARNING: re-sorting MANIFEST
> *** [MANIFEST.srt] Error code 1
> 
> make: stopped making "all" in /usr/ports/lang/perl5.36/work/perl-5.36.3
> --- lib/auto/re/re.so ---
> Unsuccessful make(ext/re): code=1536 at make_ext.pl line 584.
> *** [lib/auto/re/re.so] Error code 25
> 
> make: stopped making "all" in /usr/ports/lang/perl5.36/work/perl-5.36.3
> make: 2 errors
> 

Wow.  None of the perl5 ports build with this error.  ???

-- 
Steve

Re: perl build failure?

[David Wolfskill]

On Sat, Aug 10, 2024 at 08:05:25PM -0700, Steve Kargl wrote:
> How is perl 5.36.3 the default version when it dies with
> 
> WARNING: re-sorting MANIFEST
> *** [MANIFEST.srt] Error code 1
> 
> make: stopped making "all" in /usr/ports/lang/perl5.36/work/perl-5.36.3
> --- lib/auto/re/re.so ---
> Unsuccessful make(ext/re): code=1536 at make_ext.pl line 584.
> *** [lib/auto/re/re.so] Error code 25
> 

I do not seem to be able to reproduce this.

I already had built & installed lang/perl5.36; as an experiment, then, I:

* cd /usr/ports/
* sudo make -C lang/perl5.36
  which was uneventful, ending in:
...
> Compressing man pages (compress-man)
===>   Installing ldconfig configuration file

* echo $? (reported 0).

This was on a laptop running stable/14-n268434-5c4fd0b523f7, with the
ports tree at main-n673933-bd6e35dc25c3.

In addition, at work, we include lang/perl5.36 in the images we build
for the appliances we support; those run (a modified) FreeBSD head
(presently based on main-n271321-9ae91f59c500), with a ports tree at
main-n672216-4ff70f6df83e; these images are generally built several
times per day -- and I don't recall being aware of any reported failures
building perl5.36.  (The primary focus of my position is to look for
unexpected anomalous behavior in those images before we hand them over
for deployment; I like to think I'd be aware if such a thing were
happening.)

Perhaps some additional information for reproducing the problem might be
useful.

Peace,
david
-- 
David H. Wolfskill  da...@catwhisker.org
Someone isn't exactly courting the "childless cat lady" vote.

See https://www.catwhisker.org/~david/publickey.gpg for my public key.


signature.asc
Description: PGP signature

Re: perl build failure?

[Steve Kargl]

On Sat, Aug 10, 2024 at 08:31:19PM -0700, David Wolfskill wrote:
> On Sat, Aug 10, 2024 at 08:05:25PM -0700, Steve Kargl wrote:
> > How is perl 5.36.3 the default version when it dies with
> > 
> > WARNING: re-sorting MANIFEST
> > *** [MANIFEST.srt] Error code 1
> > 
> > make: stopped making "all" in /usr/ports/lang/perl5.36/work/perl-5.36.3
> > --- lib/auto/re/re.so ---
> > Unsuccessful make(ext/re): code=1536 at make_ext.pl line 584.
> > *** [lib/auto/re/re.so] Error code 25
> > 
> 
> I do not seem to be able to reproduce this.
> 
> I already had built & installed lang/perl5.36; as an experiment, then, I:
> 
> * cd /usr/ports/
> * sudo make -C lang/perl5.36
>   which was uneventful, ending in:
> ...
> > Compressing man pages (compress-man)
> ===>   Installing ldconfig configuration file
> 
> * echo $? (reported 0).
> 
> This was on a laptop running stable/14-n268434-5c4fd0b523f7, with the
> ports tree at main-n673933-bd6e35dc25c3.
> 
> In addition, at work, we include lang/perl5.36 in the images we build
> for the appliances we support; those run (a modified) FreeBSD head
> (presently based on main-n271321-9ae91f59c500), with a ports tree at
> main-n672216-4ff70f6df83e; these images are generally built several
> times per day -- and I don't recall being aware of any reported failures
> building perl5.36.  (The primary focus of my position is to look for
> unexpected anomalous behavior in those images before we hand them over
> for deployment; I like to think I'd be aware if such a thing were
> happening.)
> 
> Perhaps some additional information for reproducing the problem might be
> useful.
> 

This issue is occurring after a src/UPDATING buildworld/buildkernel
dance with top of tree freebsd, where I'm updating a circa Feb 2024
system.  Both source and ports were updated yesterday.  It takes a 
day to rebuild world/kernel (love llvm!!)

% cd /usr/ports/lang/perl5.36
% env MAKE_JOBS_UNSAFE=yes make
...
Making header files for XS...
Use of uninitialized value $tab[2] in substitution (s///) at ./mkheader line 
122.
Use of uninitialized value $_[0] in pattern match (m//) at ./mkheader line 108.
Illegal hexadecimal digit '!' ignored at ./mkheader line 124.
Use of uninitialized value $tab[1] in string eq at ./mkheader line 125.
Weird Canonical Decomposition of U+ at ./mkheader line 155.
require ./mkheader called at Makefile.PL line 13
Unsuccessful Makefile.PL(dist/Unicode-Normalize): code=65280 at make_ext.pl 
line 532.
*** Error code 2



-- 
Steve

UPDATING providing misinformation

[Steve Kargl]

Updating a circ Feb 2024 system with python 3.9.
Following ports/UPDATING

  # sh
  # portmaster -o lang/python311 python39
  # REINSTALL="$(pkg info -oq '*py39*')"
  # pkg delete -f "*py39*"
  # portmaster $REINSTALL
  # REBUILD=$(pkg query -g "%n:%dn" '*' | grep py3 | grep -v py311 | cut -d : 
-f 1 | sort -u)
  # portmaster $REBUILD
  # REBUILD2=$(pkg list | grep python-39 | xargs pkg which | awk '{print $6}' | 
sort -u)
  # portmaster $REBUILD2

The 'portmaster $REINSTALL' step leads to

===>>> /usr/ports/devel/py-setuptools_scm does not exist
===>>> Aborting update

% cd /usr/ports/devel
% ls | grep py-setup
py-setuptools/
py-setuptools-declarative-requirements/
py-setuptools-dso/
py-setuptools-gettext/
py-setuptools-git/
py-setuptools-git-versioning/
py-setuptools-pkg/
py-setuptools-rust/
py-setuptools-scm/

WTF?

-- 
Steve