From @jkeenan

Recently we've been getting smoke-test failures on FreeBSD-13-CURRENT in
t/perf/benchmarks.t. See, for example​:

http​://perl5.test-smoke.org/report/74129

Note that the failures in t/perf/benchmarks.t are only occuring on
threaded builds. Note also the "No plan found in TAP output" message,
which is usually associated with a premature exit to the test process
rather than failures in individual tests.

I have not seen such failures on FreeBSD-11 (or any other platform, for
that matter) and we're not getting them in FreeBSD-12 smoke test
reports. Moreover, the test file and its data in t/perf/benchmarks have
been very stable for the last twelve months.

So these failures could be due to some combination of (a) changes in
Perl 5 blead; (b) changes in FreeBSD; or (c) limitations in the
particular FreeBSD instance in which the smoking is being performed.

To diagnose this I installed a FreeBSD-13 VM on a FreeBSD-11 host -- the
same host on which we have a different FreeBSD VM for the monthly
"test-against-dev" analysis. That VM comes from the list of VirtualBox
boxes provided by Hashicorp at vagrantup.com. I suspect that this VM is
quite similar to the one from which we're getting the reports of
failures in t/perf/benchmark.t.

I have been able to reproduce the failures when I build blead with these
configuration arguments​:

#####
$ ./perl -Ilib -V​:config_args
config_args='-des -Dusedevel -Duseithreads -Doptimize=-O2 -pipe
-fstack-protector -fno-strict-aliasing';

$ cd t;./perl harness -v perf/benchmarks.t; cd -
Cannot allocate memory while trying to read 'perf/benchmarks' at
perf/benchmarks.t line 19.
Dubious, test returned 12 (wstat 3072, 0xc00)
No subtests run

Test Summary Report

perf/benchmarks.t (Wstat​: 3072 Tests​: 0 Failed​: 0)
  Non-zero exit status​: 12
  Parse errors​: No plan found in TAP output
Files=1, Tests=0, 0 wallclock secs ( 0.01 usr 0.01 sys + 0.01 cusr
0.01 csys = 0.03 CPU)
Result​: FAIL
#####

However, when I configure without threads, all tests PASS​:

#####
[perl2] $ ./perl -Ilib -V​:config_args
config_args='-des -Dusedevel';
[perl2] $ cd t;./perl harness perf/benchmarks.t; cd -
perf/benchmarks.t .. ok
All tests successful.
Files=1, Tests=1728, 1 wallclock secs ( 0.18 usr 0.02 sys + 0.08 cusr
  0.03 csys = 0.30 CPU)
Result​: PASS
/home/jkeenan/gitwork/perl2
#####

That reproduces the results Carlos Guevara is reporting in the smoke
test report cited above.

When I conduct an internet search for "Cannot allocate memory while
trying to read", I mostly come up with cases on Linux. There's little
that's FreeBSD-specific. For what it's worth, here are two lines from
the output of 'ulimit -a' on this VM​:

#####
max user processes (-u) 3416
open files (-n) 13635
#####

In contrast, on the host on which this VM sits, the corresponding lines
from 'ulimit -a' are​:

#####
max user processes (-u) 12171
open files (-n) 22500
#####

This leads me to suspect that the problem is (c) -- a resource
constraint in the way this particular "box" has been set up -- rather
than something wrong in blead or in FreeBSD-13.

I experimented at splitting the fixtures in t/perf/benchmarks into two
files, then modifying t/perf/benchmarks.t so that it does one run over
each of the two files. That worked; I did not get the "Cannot allocate
memory" error.

Unless someone has a better way of handling this problem, I'd like to
apply that splitting-up to blead so that we can reduce the amount of
"noise" coming from these apparently spurious test failures.

Thank you very much.
Jim Keenan

From @jkeenan

Summary of my perl5 (revision 5 version 29 subversion 5) configuration​:
  Commit id​: 3db0bcc
  Platform​:
  osname=freebsd
  osvers=13.0-current
  archname=amd64-freebsd-thread-multi
  uname='freebsd perl-reporter-05 13.0-current freebsd 13.0-current r340361 generic amd64 '
  config_args='-des -Dusedevel -Duseithreads -Doptimize=-O2 -pipe -fstack-protector -fno-strict-aliasing'
  hint=recommended
  useposix=true
  d_sigaction=define
  useithreads=define
  usemultiplicity=define
  use64bitint=define
  use64bitall=define
  uselongdouble=undef
  usemymalloc=n
  default_inc_excludes_dot=define
  bincompat5005=undef
  Compiler​:
  cc='cc'
  ccflags ='-DHAS_FPSETMASK -DHAS_FLOATINGPOINT_H -fno-strict-aliasing -pipe -fstack-protector-strong -I/usr/local/include -D_FORTIFY_SOURCE=2'
  optimize='-O2 -pipe -fstack-protector -fno-strict-aliasing'
  cppflags='-DHAS_FPSETMASK -DHAS_FLOATINGPOINT_H -fno-strict-aliasing -pipe -fstack-protector-strong -I/usr/local/include'
  ccversion=''
  gccversion='4.2.1 Compatible FreeBSD Clang 6.0.1 (tags/RELEASE_601/final 335540)'
  gccosandvers=''
  intsize=4
  longsize=8
  ptrsize=8
  doublesize=8
  byteorder=12345678
  doublekind=3
  d_longlong=define
  longlongsize=8
  d_longdbl=define
  longdblsize=16
  longdblkind=3
  ivtype='long'
  ivsize=8
  nvtype='double'
  nvsize=8
  Off_t='off_t'
  lseeksize=8
  alignbytes=8
  prototype=define
  Linker and Libraries​:
  ld='cc'
  ldflags ='-pthread -Wl,-E -fstack-protector-strong -L/usr/local/lib'
  libpth=/usr/lib /usr/local/lib /usr/lib/clang/6.0.1/lib /usr/lib
  libs=-lpthread -lgdbm -ldl -lm -lcrypt -lutil
  perllibs=-lpthread -ldl -lm -lcrypt -lutil
  libc=
  so=so
  useshrplib=false
  libperl=libperl.a
  gnulibc_version=''
  Dynamic Linking​:
  dlsrc=dl_dlopen.xs
  dlext=so
  d_dlsymun=undef
  ccdlflags=' '
  cccdlflags='-DPIC -fPIC'
  lddlflags='-shared -L/usr/local/lib -fstack-protector-strong'

Characteristics of this binary (from libperl)​:
  Compile-time options​:
  HAS_TIMES
  MULTIPLICITY
  PERLIO_LAYERS
  PERL_COPY_ON_WRITE
  PERL_DONT_CREATE_GVSV
  PERL_IMPLICIT_CONTEXT
  PERL_MALLOC_WRAP
  PERL_OP_PARENT
  PERL_PRESERVE_IVUV
  PERL_USE_DEVEL
  USE_64_BIT_ALL
  USE_64_BIT_INT
  USE_ITHREADS
  USE_LARGE_FILES
  USE_LOCALE
  USE_LOCALE_COLLATE
  USE_LOCALE_CTYPE
  USE_LOCALE_NUMERIC
  USE_LOCALE_TIME
  USE_PERLIO
  USE_PERL_ATOF
  USE_REENTRANT_API
  Built under freebsd
  Compiled at Nov 17 2018 14​:26​:37
  %ENV​:
  PERL2DIR="/home/jkeenan/gitwork/perl2"
  PERL_WORKDIR="/home/jkeenan/gitwork/perl"
  @​INC​:
  lib
  /usr/local/lib/perl5/site_perl/5.29.5/amd64-freebsd-thread-multi
  /usr/local/lib/perl5/site_perl/5.29.5
  /usr/local/lib/perl5/5.29.5/amd64-freebsd-thread-multi
  /usr/local/lib/perl5/5.29.5

From @iabyn

On Sat, Nov 17, 2018 at 12​:58​:47PM -0800, James E Keenan (via RT) wrote​:

Recently we've been getting smoke-test failures on FreeBSD-13-CURRENT in
t/perf/benchmarks.t. See, for example​:

$ cd t;./perl harness -v perf/benchmarks.t; cd -
Cannot allocate memory while trying to read 'perf/benchmarks' at
perf/benchmarks.t line 19.

At that point perf/benchmarks.t is simply executing

  do 'perf/benchmarks';

and perf/benchmarks is a 2000-ish line chunk of perl src which
creates an array of 864 hashes.

So either the VM doesn't have enough memory to compile that bit of code,
or its hitting a stack limit while recursively scanning the op tree during
compilation. In which case...

max user processes (-u) 3416
open files (-n) 13635

it's much more likely to be one of these​:

  data seg size (kbytes, -d) unlimited
  max memory size (kbytes, -m) unlimited
  stack size (kbytes, -s) 8192
  virtual memory (kbytes, -v) unlimited

What do those give you on the BSD VM and its host?

I experimented at splitting the fixtures in t/perf/benchmarks into two
files, then modifying t/perf/benchmarks.t so that it does one run over
each of the two files. That worked; I did not get the "Cannot allocate
memory" error.

Unless someone has a better way of handling this problem, I'd like to
apply that splitting-up to blead so that we can reduce the amount of
"noise" coming from these apparently spurious test failures.

The file t/perf/benchmarks is the core of a benchmarking system for the
perl core; t/perf/benchmarks.t is just a quick sanity check that the file
is compilabe and that the code snippets within it are compilable.
Splitting t/perf/benchmarks in two involves changing a lot more
infrastructure than just t/perf/benchmarks.t.

We should probably just skip the test file when 'ulimit -X' gives a low
value (for whatever X it turns out to be) - and if possible find out
why perl is choking on a 2000-line src file, and maybe fix some of the
recursion limits in op.c if that is indeed the problem.

--
Counsellor Troi states something other than the blindingly obvious.
  -- Things That Never Happen in "Star Trek" #16

The RT System itself - Status changed from 'new' to 'open'

From @jkeenan

On Sat, 17 Nov 2018 22​:37​:12 GMT, davem wrote​:

On Sat, Nov 17, 2018 at 12​:58​:47PM -0800, James E Keenan (via RT) wrote​:

Recently we've been getting smoke-test failures on FreeBSD-13-CURRENT in
t/perf/benchmarks.t. See, for example​:

$ cd t;./perl harness -v perf/benchmarks.t; cd -
Cannot allocate memory while trying to read 'perf/benchmarks' at
perf/benchmarks.t line 19.

At that point perf/benchmarks.t is simply executing

do 'perf/benchmarks';

and perf/benchmarks is a 2000-ish line chunk of perl src which
creates an array of 864 hashes.

So either the VM doesn't have enough memory to compile that bit of code,
or its hitting a stack limit while recursively scanning the op tree during
compilation. In which case...

max user processes (-u) 3416
open files (-n) 13635

it's much more likely to be one of these​:

data seg size           \(kbytes\, \-d\) unlimited
max memory size         \(kbytes\, \-m\) unlimited
stack size              \(kbytes\, \-s\) 8192
virtual memory          \(kbytes\, \-v\) unlimited

What do those give you on the BSD VM and its host?

# On the host​:

#####
$ ulimit -a | sort
core file size (512-blocks, -c) unlimited
cpu time (seconds, -t) unlimited
data seg size (kbytes, -d) 33554432
file size (512-blocks, -f) unlimited
kqueues (-k) unlimited
locked memory (kbytes, -l) 64
max memory size (kbytes, -m) unlimited
max user processes (-u) 12171
open files (-n) 22500
pseudo-terminals (-p) unlimited
socket buffer size (bytes, -b) unlimited
stack size (kbytes, -s) 524288
swap limit (kbytes, -w) unlimited
umtx shared locks (-o) unlimited
virtual mem size (kbytes, -v) unlimited
#####

In the FreeBSD-13 VM​:

#####
% ulimit -a | sort
core file size (512-blocks, -c) unlimited
cpu time (seconds, -t) unlimited
data seg size (kbytes, -d) 33554432
file size (512-blocks, -f) unlimited
kqueues (-k) unlimited
locked memory (kbytes, -l) 64
max memory size (kbytes, -m) unlimited
max user processes (-u) 3416
open files (-n) 13635
pseudo-terminals (-p) unlimited
socket buffer size (bytes, -b) unlimited
stack size (kbytes, -s) 524288
swap limit (kbytes, -w) unlimited
umtx shared locks (-o) unlimited
virtual mem size (kbytes, -v) unlimited
#####

I experimented at splitting the fixtures in t/perf/benchmarks into two
files, then modifying t/perf/benchmarks.t so that it does one run over
each of the two files. That worked; I did not get the "Cannot allocate
memory" error.

Unless someone has a better way of handling this problem, I'd like to
apply that splitting-up to blead so that we can reduce the amount of
"noise" coming from these apparently spurious test failures.

The file t/perf/benchmarks is the core of a benchmarking system for the
perl core; t/perf/benchmarks.t is just a quick sanity check that the file
is compilabe and that the code snippets within it are compilable.
Splitting t/perf/benchmarks in two involves changing a lot more
infrastructure than just t/perf/benchmarks.t.

We should probably just skip the test file when 'ulimit -X' gives a low
value (for whatever X it turns out to be) - and if possible find out
why perl is choking on a 2000-line src file, and maybe fix some of the
recursion limits in op.c if that is indeed the problem.

--
James E Keenan (jkeenan@​cpan.org)

From @eserte

Dana Sat, 17 Nov 2018 12​:58​:46 -0800, jkeenan@​pobox.com reče​:

Recently we've been getting smoke-test failures on FreeBSD-13-CURRENT in
t/perf/benchmarks.t. See, for example​:

http​://perl5.test-smoke.org/report/74129

Note that the failures in t/perf/benchmarks.t are only occuring on
threaded builds. Note also the "No plan found in TAP output" message,
which is usually associated with a premature exit to the test process
rather than failures in individual tests.

I have not seen such failures on FreeBSD-11 (or any other platform, for
that matter) and we're not getting them in FreeBSD-12 smoke test
reports. Moreover, the test file and its data in t/perf/benchmarks have
been very stable for the last twelve months.

So these failures could be due to some combination of (a) changes in
Perl 5 blead; (b) changes in FreeBSD; or (c) limitations in the
particular FreeBSD instance in which the smoking is being performed.

To diagnose this I installed a FreeBSD-13 VM on a FreeBSD-11 host -- the
same host on which we have a different FreeBSD VM for the monthly
"test-against-dev" analysis. That VM comes from the list of VirtualBox
boxes provided by Hashicorp at vagrantup.com. I suspect that this VM is
quite similar to the one from which we're getting the reports of
failures in t/perf/benchmark.t.

I have been able to reproduce the failures when I build blead with these
configuration arguments​:

#####
$ ./perl -Ilib -V​:config_args
config_args='-des -Dusedevel -Duseithreads -Doptimize=-O2 -pipe
-fstack-protector -fno-strict-aliasing';

$ cd t;./perl harness -v perf/benchmarks.t; cd -
Cannot allocate memory while trying to read 'perf/benchmarks' at
perf/benchmarks.t line 19.
Dubious, test returned 12 (wstat 3072, 0xc00)
No subtests run

Test Summary Report
-------------------
perf/benchmarks.t (Wstat​: 3072 Tests​: 0 Failed​: 0)
Non-zero exit status​: 12
Parse errors​: No plan found in TAP output
Files=1, Tests=0, 0 wallclock secs ( 0.01 usr 0.01 sys + 0.01 cusr
0.01 csys = 0.03 CPU)
Result​: FAIL
#####

However, when I configure without threads, all tests PASS​:

#####
[perl2] $ ./perl -Ilib -V​:config_args
config_args='-des -Dusedevel';
[perl2] $ cd t;./perl harness perf/benchmarks.t; cd -
perf/benchmarks.t .. ok
All tests successful.
Files=1, Tests=1728, 1 wallclock secs ( 0.18 usr 0.02 sys + 0.08 cusr
0.03 csys = 0.30 CPU)
Result​: PASS
/home/jkeenan/gitwork/perl2
#####

That reproduces the results Carlos Guevara is reporting in the smoke
test report cited above.

When I conduct an internet search for "Cannot allocate memory while
trying to read", I mostly come up with cases on Linux. There's little
that's FreeBSD-specific. For what it's worth, here are two lines from
the output of 'ulimit -a' on this VM​:

#####
max user processes (-u) 3416
open files (-n) 13635
#####

In contrast, on the host on which this VM sits, the corresponding lines
from 'ulimit -a' are​:

#####
max user processes (-u) 12171
open files (-n) 22500
#####

This leads me to suspect that the problem is (c) -- a resource
constraint in the way this particular "box" has been set up -- rather
than something wrong in blead or in FreeBSD-13.

I experimented at splitting the fixtures in t/perf/benchmarks into two
files, then modifying t/perf/benchmarks.t so that it does one run over
each of the two files. That worked; I did not get the "Cannot allocate
memory" error.

Unless someone has a better way of handling this problem, I'd like to
apply that splitting-up to blead so that we can reduce the amount of
"noise" coming from these apparently spurious test failures.

Thank you very much.
Jim Keenan

Maybe related​: ZEFRAM/XML-Easy-0.011.tar.gz fails with perl 5.29.4 with a warning

  WARNING​: Complex regular subexpression recursion limit (65534) exceeded at t/syntax_main.t line 79, <GEN0> line 1635.

(seen on linux + freebsd). This does not happen with 5.29.3. So maybe something in blead is consuming more resources?

From @jkeenan

On Sun, 18 Nov 2018 16​:52​:44 GMT, slaven@​rezic.de wrote​:

Dana Sat, 17 Nov 2018 12​:58​:46 -0800, jkeenan@​pobox.com reče​:

Recently we've been getting smoke-test failures on FreeBSD-13-CURRENT
in
t/perf/benchmarks.t. See, for example​:

http​://perl5.test-smoke.org/report/74129

Note that the failures in t/perf/benchmarks.t are only occuring on
threaded builds. Note also the "No plan found in TAP output"
message,
which is usually associated with a premature exit to the test process
rather than failures in individual tests.

I have not seen such failures on FreeBSD-11 (or any other platform,
for
that matter) and we're not getting them in FreeBSD-12 smoke test
reports. Moreover, the test file and its data in t/perf/benchmarks
have
been very stable for the last twelve months.

So these failures could be due to some combination of (a) changes in
Perl 5 blead; (b) changes in FreeBSD; or (c) limitations in the
particular FreeBSD instance in which the smoking is being performed.

To diagnose this I installed a FreeBSD-13 VM on a FreeBSD-11 host --
the
same host on which we have a different FreeBSD VM for the monthly
"test-against-dev" analysis. That VM comes from the list of
VirtualBox
boxes provided by Hashicorp at vagrantup.com. I suspect that this VM
is
quite similar to the one from which we're getting the reports of
failures in t/perf/benchmark.t.

I have been able to reproduce the failures when I build blead with
these
configuration arguments​:

#####
$ ./perl -Ilib -V​:config_args
config_args='-des -Dusedevel -Duseithreads -Doptimize=-O2 -pipe
-fstack-protector -fno-strict-aliasing';

$ cd t;./perl harness -v perf/benchmarks.t; cd -
Cannot allocate memory while trying to read 'perf/benchmarks' at
perf/benchmarks.t line 19.
Dubious, test returned 12 (wstat 3072, 0xc00)
No subtests run

Test Summary Report
-------------------
perf/benchmarks.t (Wstat​: 3072 Tests​: 0 Failed​: 0)
Non-zero exit status​: 12
Parse errors​: No plan found in TAP output
Files=1, Tests=0, 0 wallclock secs ( 0.01 usr 0.01 sys + 0.01
cusr
0.01 csys = 0.03 CPU)
Result​: FAIL
#####

However, when I configure without threads, all tests PASS​:

#####
[perl2] $ ./perl -Ilib -V​:config_args
config_args='-des -Dusedevel';
[perl2] $ cd t;./perl harness perf/benchmarks.t; cd -
perf/benchmarks.t .. ok
All tests successful.
Files=1, Tests=1728, 1 wallclock secs ( 0.18 usr 0.02 sys + 0.08
cusr
0.03 csys = 0.30 CPU)
Result​: PASS
/home/jkeenan/gitwork/perl2
#####

That reproduces the results Carlos Guevara is reporting in the smoke
test report cited above.

When I conduct an internet search for "Cannot allocate memory while
trying to read", I mostly come up with cases on Linux. There's
little
that's FreeBSD-specific. For what it's worth, here are two lines
from
the output of 'ulimit -a' on this VM​:

#####
max user processes (-u) 3416
open files (-n) 13635
#####

In contrast, on the host on which this VM sits, the corresponding
lines
from 'ulimit -a' are​:

#####
max user processes (-u) 12171
open files (-n) 22500
#####

This leads me to suspect that the problem is (c) -- a resource
constraint in the way this particular "box" has been set up -- rather
than something wrong in blead or in FreeBSD-13.

I experimented at splitting the fixtures in t/perf/benchmarks into
two
files, then modifying t/perf/benchmarks.t so that it does one run
over
each of the two files. That worked; I did not get the "Cannot
allocate
memory" error.

Unless someone has a better way of handling this problem, I'd like to
apply that splitting-up to blead so that we can reduce the amount of
"noise" coming from these apparently spurious test failures.

Thank you very much.
Jim Keenan

Maybe related​: ZEFRAM/XML-Easy-0.011.tar.gz fails with perl 5.29.4
with a warning

WARNING​: Complex regular subexpression recursion limit (65534)
exceeded at t/syntax_main.t line 79, <GEN0> line 1635.

(seen on linux + freebsd). This does not happen with 5.29.3. So maybe
something in blead is consuming more resources?

No, this is a situation where a change in blead exposed sub-optimal code in XML​::Easy's test suite. This was detected in the October "CPAN-river-3000" run and I provided a patch to the author. See​: https://rt.cpan.org/Ticket/Display.html?id=127416#txn-1816119

Thank you very much.
--
James E Keenan (jkeenan@​cpan.org)

From @iabyn

On Sat, Nov 17, 2018 at 02​:44​:15PM -0800, James E Keenan via RT wrote​:

On Sat, 17 Nov 2018 22​:37​:12 GMT, davem wrote​:

On Sat, Nov 17, 2018 at 12​:58​:47PM -0800, James E Keenan (via RT) wrote​:

Recently we've been getting smoke-test failures on FreeBSD-13-CURRENT in
t/perf/benchmarks.t. See, for example​:

$ cd t;./perl harness -v perf/benchmarks.t; cd -
Cannot allocate memory while trying to read 'perf/benchmarks' at
perf/benchmarks.t line 19.

At that point perf/benchmarks.t is simply executing

do 'perf/benchmarks';

and perf/benchmarks is a 2000-ish line chunk of perl src which
creates an array of 864 hashes.

So either the VM doesn't have enough memory to compile that bit of code,
or its hitting a stack limit while recursively scanning the op tree during
compilation. In which case...

max user processes (-u) 3416
open files (-n) 13635

it's much more likely to be one of these​:

data seg size           \(kbytes\, \-d\) unlimited
max memory size         \(kbytes\, \-m\) unlimited
stack size              \(kbytes\, \-s\) 8192
virtual memory          \(kbytes\, \-v\) unlimited

What do those give you on the BSD VM and its host?

[ all other values identical on host and VM]

That's odd.

Does it fail if you execute the script directly​:

  ./perl t/perf/benchmarks

and if so does it appear to be just the size of the array of hashes it
creates, or something specific about the content?

--
Monto Blanco... scorchio!

From @jkeenan

On Mon, 19 Nov 2018 14​:55​:04 GMT, davem wrote​:

On Sat, Nov 17, 2018 at 02​:44​:15PM -0800, James E Keenan via RT wrote​:

On Sat, 17 Nov 2018 22​:37​:12 GMT, davem wrote​:

On Sat, Nov 17, 2018 at 12​:58​:47PM -0800, James E Keenan (via RT)
wrote​:

Recently we've been getting smoke-test failures on FreeBSD-13-
CURRENT in
t/perf/benchmarks.t. See, for example​:

$ cd t;./perl harness -v perf/benchmarks.t; cd -
Cannot allocate memory while trying to read 'perf/benchmarks' at
perf/benchmarks.t line 19.

At that point perf/benchmarks.t is simply executing

do 'perf/benchmarks';

and perf/benchmarks is a 2000-ish line chunk of perl src which
creates an array of 864 hashes.

So either the VM doesn't have enough memory to compile that bit of
code,
or its hitting a stack limit while recursively scanning the op tree
during
compilation. In which case...

max user processes (-u) 3416
open files (-n) 13635

it's much more likely to be one of these​:

data seg size (kbytes, -d) unlimited
max memory size (kbytes, -m) unlimited
stack size (kbytes, -s) 8192
virtual memory (kbytes, -v) unlimited

What do those give you on the BSD VM and its host?

[ all other values identical on host and VM]

That's odd.

Does it fail if you execute the script directly​:

./perl t/perf/benchmarks

and if so does it appear to be just the size of the array of hashes it
creates, or something specific about the content?

#####
[perl] $ ./perl -Ilib -v | head -2 | tail -1
This is perl 5, version 29, subversion 5 (v5.29.5 (v5.29.4-91-g2d0e7d1fca)) built for amd64-freebsd-thread-multi
[perl] $ ./perl -Ilib -V​:config_args
config_args='-des -Dusedevel -Duseithreads -Doptimize=-O2 -pipe -fstack-protector -fno-strict-aliasing';
[perl] $ cd t;./perl harness -v perf/benchmarks.t; cd -
Cannot allocate memory while trying to read 'perf/benchmarks' at perf/benchmarks.t line 19.
Dubious, test returned 12 (wstat 3072, 0xc00)
No subtests run

Test Summary Report

perf/benchmarks.t (Wstat​: 3072 Tests​: 0 Failed​: 0)
  Non-zero exit status​: 12
  Parse errors​: No plan found in TAP output
Files=1, Tests=0, 0 wallclock secs ( 0.02 usr 0.02 sys + 0.02 cusr 0.00 csys = 0.05 CPU)
Result​: FAIL
/home/jkeenan/gitwork/perl
[perl] $ ./perl -Ilib t/perf/benchmarks.t
Cannot allocate memory while trying to read 'perf/benchmarks' at t/perf/benchmarks.t line 19.
[perl] $ ./perl -Ilib t/perf/benchmarks
[perl] $
#####

Based on my debugging the other day, I would say that it's the size of the array of hashes rather than any specific element therein.

I started by commenting out all tests, then uncommenting them one "block" at a time -- where a block is a group of elements with similar keys -- then running the test. Repeat until FAIL. I got about 70% of the way through the file before I got the error.

I then reversed the process, commenting all tests out and uncommenting them from the bottom upwards. Once again, I got about 70% of the way through the file before the error occurred.

--
James E Keenan (jkeenan@​cpan.org)

From @iabyn

On Mon, Nov 19, 2018 at 10​:57​:48AM -0800, James E Keenan via RT wrote​:

[perl] $ ./perl -Ilib t/perf/benchmarks.t
Cannot allocate memory while trying to read 'perf/benchmarks' at t/perf/benchmarks.t line 19.
[perl] $ ./perl -Ilib t/perf/benchmarks

I've just pushed the following commit. Can you retry ./perl -Ilib t/perf/benchmarks.t and show me what the error message is now?

commit 516795a
Author​: David Mitchell <davem@​iabyn.com>
AuthorDate​: Tue Nov 20 10​:59​:22 2018 +0000

  t/perf/benchmarks.t​: improve do error checks
 
  Make the checks for "do 't/perf/benchmarks'" look more like those
  suggested for 'do' in perlfunc.
 
  In particular, this may help track down the issue in RT #133663.

--
That he said that that that that is is is debatable, is debatable.

From @jkeenan

On Tue, 20 Nov 2018 11​:12​:39 GMT, davem wrote​:

On Mon, Nov 19, 2018 at 10​:57​:48AM -0800, James E Keenan via RT wrote​:

[perl] $ ./perl -Ilib t/perf/benchmarks.t
Cannot allocate memory while trying to read 'perf/benchmarks' at
t/perf/benchmarks.t line 19.
[perl] $ ./perl -Ilib t/perf/benchmarks

I've just pushed the following commit. Can you retry ./perl -Ilib
t/perf/benchmarks.t and show me what the error message is now?

commit 516795a
Author​: David Mitchell <davem@​iabyn.com>
AuthorDate​: Tue Nov 20 10​:59​:22 2018 +0000

t/perf/benchmarks.t​: improve do error checks

Make the checks for "do 't/perf/benchmarks'" look more like those
suggested for 'do' in perlfunc.

In particular, this may help track down the issue in RT #133663.

This looks good.

#####
[perl] $ uname -mrs
FreeBSD 13.0-CURRENT amd64
[perl] $ ./perl -Ilib -v | head -2 | tail -1
This is perl 5, version 29, subversion 5 (v5.29.5 (v5.29.4-99-g516795a038)) built for amd64-freebsd-thread-multi
[perl] $ ./perl -Ilib -V​:config_args
config_args='-des -Dusedevel -Duseithreads -Doptimize=-O2 -pipe -fstack-protector -fno-strict-aliasing';
[perl] $ cd t;./perl harness perf/benchmarks.t; cd -
perf/benchmarks.t .. ok
All tests successful.
Files=1, Tests=1728, 0 wallclock secs ( 0.18 usr 0.01 sys + 0.09 cusr 0.02 csys = 0.30 CPU)
Result​: PASS
/home/jkeenan/gitwork/perl
[perl] $ ./perl -Ilib t/perf/benchmarks.t
...
ok 1727 - running regex​::whilem​::max_captures_fail
ok 1728 - running regex​::whilem​::min_captures_fail

#####

I will close this ticket once I see smoke-test results. Thank you very much.

--
James E Keenan (jkeenan@​cpan.org)

./perl -Ilib t/perf/benchmarks.t

I should have checked the smoke results a year-and-a-half ago!

[perl] $ uname -mrs
FreeBSD 13.0-CURRENT amd64
[perl] $ ./perl -v | head -2 | tail -1
This is perl 5, version 31, subversion 9 (v5.31.9 (v5.31.8-86-g2b301921ff)) built for amd64-freebsd-thread-multi
[perl] $ cc --version
FreeBSD clang version 6.0.1 (tags/RELEASE_601/final 335540) (based on LLVM 6.0.1)
Target: x86_64-unknown-freebsd13.0
Thread model: posix
InstalledDir: /usr/bin
[perl] $ ./perl -Ilib -V:config_args
config_args='-des -Dusedevel -Duseithreads -Doptimize=-O2 -pipe -fstack-protector -fno-strict-aliasing';
[perl] $ cd t;./perl harness perf/benchmarks.t; cd -
perf/benchmarks.t .. ok         
All tests successful.
Files=1, Tests=1732,  0 wallclock secs ( 0.20 usr  0.00 sys +  0.09 cusr  0.02 csys =  0.31 CPU)
Result: PASS

We're good. Closing.