Regression on Cygwin: Problems with parallel make in 4.4

[Ken Brown]

Several packages that used to build fine on Cygwin with parallel make 
now fail to build.  I either get strange errors or a hang.  The problems 
started with version 4.4 and seem to stem from the fact that make now 
uses a FIFO by default instead of a pipe.  If I use the flag 
'--jobserver-style=pipe', parallel make works again.  A recent example 
is described here:


  https://tug.org/pipermail/tlbuild/2023q1/005290.html

There may be a bug in Cygwin's FIFO implementation underlying this, but 
nevertheless it's a regression from 4.3 to 4.4, and most Cygwin users 
won't know that they have to use the --jobserver-style flag.  I suggest 
that you provide a configure option to set the jobserver style at the 
time make is built, and set it to 'pipe' by default on Cygwin.


Thanks.

Ken

Re: Regression on Cygwin: Problems with parallel make in 4.4

[Ken Brown]

On 2/15/2023 1:34 PM, Bruno Haible wrote:

Paul Smith wrote:

And possibly also on GNU/Hurd. Cf.
https://lists.gnu.org/archive/html/bug-make/2023-01/msg00107.html


I may be misremembering but I thought that you had tried forcing the
pipe jobserver option on GNU/Hurd and it didn't help.


This must be a misunderstanding. I never said that. I don't even know
how to do this, within the framework of GNU make's test suite.

All I said was, comparing the Cygwin failures with the GNU/Hurd failures [1]:
   "In all these failures, where we see an error message "Leftover temporary
files" in GNU/Hurd, we see "Device or resource busy" in Cygwin."

Bruno

[1] https://lists.gnu.org/archive/html/bug-make/2023-01/msg00134.html


One thing to keep in mind here is that your tests on Cygwin were done on 
Cygwin 2.9.0, in which FIFOs were very poorly supported.  For example, a 
FIFO couldn't have multiple readers or writers.  If GNU/Hurd has similar 
limitations, that could explain the similar test results.


Cygwin's FIFO implementation was completely rewritten starting with 3.0, 
and it now supports multiple readers and writers.  So the issue with 
current Cygwin is more likely a bug in Cygwin and/or make.


Marco, did you by any chance run the test suite when you built make for 
Cygwin?  It would by interesting to know what test failures there are 
with Cygwin 3.4.6.


Ken

P.S. Here's a footnote about the failures I'm seeing.  I cited

  https://tug.org/pipermail/tlbuild/2023q1/005290.html ,

in which I reported problems with the build of TeX Live.  I said I had a 
hang in 'make -j13 world'.  This was an oversimplification.  What really 
happens is that the hang occurs when I run the 'Build' script provided 
by the TeX Live sources.  This is a shell script that runs make in a 
further sub-shell.  If I simply run 'make -j13 world' without using the 
Build script, it doesn't hang.  I don't know if this provides any clue 
as to where the problem is.

Re: Regression on Cygwin: Problems with parallel make in 4.4

[Ken Brown]

On 2/19/2023 8:49 AM, Paul Smith wrote:

On Wed, 2023-02-15 at 13:57 -0500, Ken Brown wrote:

One thing to keep in mind here is that your tests on Cygwin were done
on Cygwin 2.9.0, in which FIFOs were very poorly supported.  For
example, a FIFO couldn't have multiple readers or writers.  If
GNU/Hurd has similar limitations, that could explain the similar test
results.

Cygwin's FIFO implementation was completely rewritten starting with
3.0, and it now supports multiple readers and writers.  So the issue
with current Cygwin is more likely a bug in Cygwin and/or make.


GNU make _definitely_ has multiple readers/writers (that's really the
whole point :)).

So I'm not sure where to go from here on Cygwin.  Should I force Cygwin
builds to use the "pipe" jobserver, as I've done for GNU/Hurd?


My preference would be for you to provide a configure option, which 
defaults to "pipe" on Cygwin.  But if you don't want to do that, then I 
think you should force Cygwin builds to use "pipe".



 Should
I add preprocessor checks (or autoconf checks?) to try to figure out
which version of Cygwin is being used and choose based on that?  I
don't know how to check versions.


That wouldn't help.  There's currently no known version of Cygwin on 
which the fifo option works well.  (Bruno tested on old version, I 
tested the current version.)


Ken

Re: Regression on Cygwin: Problems with parallel make in 4.4

[Ken Brown]

On 2/19/2023 9:29 AM, Paul Smith wrote:

On Sun, 2023-02-19 at 09:17 -0500, Ken Brown wrote:

So I'm not sure where to go from here on Cygwin.  Should I force
Cygwin builds to use the "pipe" jobserver, as I've done for
GNU/Hurd?


My preference would be for you to provide a configure option, which
defaults to "pipe" on Cygwin.  But if you don't want to do that, then
I think you should force Cygwin builds to use "pipe".


There is (in my current implementation, not released yet) a way to do
that but it's a little obscure; you need to use:

  ./configure CPPFLAGS=-DJOBSERVER_USE_FIFO=0

or

  ./configure CPPFLAGS=-DJOBSERVER_USE_FIFO=1


That's good.


If you set it explicitly then that setting is obeyed regardless of the
capabilities of the system (well, if mkfifo() is not available then
trying to enable it is a no-op of course).


Should I add preprocessor checks (or autoconf checks?) to try to
figure out which version of Cygwin is being used and choose based
on that?  I don't know how to check versions.


That wouldn't help.  There's currently no known version of Cygwin on
which the fifo option works well.  (Bruno tested on old version, I
tested the current version.)


OK but I noted you said:


If I simply run 'make -j13 world' without using the Build script, it
doesn't hang.


so I wasn't sure if we were considering that a "full failure".


I agree that it's not a full failure, but it's annoying when it happens. 
 In my use case (building TeX Live using the supplied Build script), 
the build runs for about 2 hours before it fails.



 I guess
if we don't know why it happens, we just have to disable mkfifo().

I will change the default setting of the jobserver to use "pipe" on
Cygwin, at least for now.


Thanks.

Ken

Re: GNU Make 4.4.0.91 on Cygwin

[Ken Brown]

On 2/19/2023 2:59 PM, Paul Smith wrote:

On Sun, 2023-02-19 at 20:32 +0100, Bruno Haible wrote:

All "Device or resource busy" failures are gone. Only the 1 failure
in category 'misc/general4' is still present.


Hm.  This is a test of this:
https://savannah.gnu.org/bugs/index.php?57674

which explicitly removes PATH from the environment completely.  It
should use a default PATH, which should mean "echo" works even so.  I'm
not really sure how this should react on a cygwin system, but I guess
since it behaves properly on Windows and POSIX it should work on cygwin
as well.  I'll have to figure out why only cygwin isn't working.


All tests pass on my Cygwin 3.4.6 system.

Ken

Re: Regression on Cygwin: Problems with parallel make in 4.4

[Ken Brown]

On 2/19/2023 9:29 AM, Paul Smith wrote:

I will change the default setting of the jobserver to use "pipe" on
Cygwin, at least for now.


Parallel make is still not working reliably.  I've just discovered that 
my TeX Live build logs have several occurrences of the following warning:


  jobserver unavailable: using -j1.  Add '+' to parent make rule.

This has been going on since I first started using 4.4.0.90 with the 
jobserver style set to "pipe".  It *only* occurs when the jobserver uses 
a pipe.  Two others in the Cygwin community are reporting the same thing 
when they build projects using cmake (with Unix Makefiles and make 
4.4.0.91).


I don't know if this is Cygwin-specific, since other platforms (except 
GNU/Hurd) are using fifos.  So the "pipe" jobserver style might not have 
gotten a lot of testing since the switch to fifos was made.


I guess the first step in tracking this down is for people working on 
other platforms to see if they can reproduce the problem when they set 
the jobserver style to "pipe".


Unfortunately, I don't have a simple reproduction recipe.  I myself have 
seen it when building TeX Live, as I've said, and also when building 
doxygen (which uses cmake with Unix Makefiles).  The other Cygwin users 
who have reported the problem (both CC'd) haven't yet said what cmake 
projects they were building.


Ken

Re: Regression on Cygwin: Problems with parallel make in 4.4

[Ken Brown]

On 2/21/2023 12:27 AM, Paul Smith wrote:

On Mon, 2023-02-20 at 20:21 -0500, Ken Brown wrote:

Parallel make is still not working reliably.  I've just discovered
that my TeX Live build logs have several occurrences of the following
warning:

    jobserver unavailable: using -j1.  Add '+' to parent make rule.

This has been going on since I first started using 4.4.0.90 with the
jobserver style set to "pipe".  It *only* occurs when the jobserver
uses a pipe.  Two others in the Cygwin community are reporting the
same thing when they build projects using cmake (with Unix Makefiles
and make 4.4.0.91).


Just to note, I do run the regression test suite with the equivalent of
"jobserver-style=pipe" (basically I force the configure to believe that
the system doesn't support mkfifo() so it falls back to "pipe" mode).

Can you run builds with GNU Make 4.4 and --jobserver-style=pipe?  Do
you see the same warnings there?  I just want to know if the problem
happened in 4.4 or post-4.4.  If the latter, that's very odd.  If the
former, it's more understandable because we did make the "pipe" mode
more strict in 4.4.


I do see it in 4.4 with --jobserver-style=pipe.  I misspoke when I said 
I first saw it in 4.4.0.90.



This message means only one thing: the parent make didn't provide the
proper environment (that is, the correct file descriptors) to its sub-
make.  However, there could be different ways in which the environment
is wrong.

If you have the ability, it would be helpful if you could apply the
attached patch to your build of GNU Make and try the build again, and
let me know which of these different messages are generated right
before the above error.


I'll be able to do that later today.


It would also be very helpful if someone could provide me with the
complete rule that the parent make uses to invoke the sub-make, where
that message appears.  And, if you can include the output that make
generates to invoke the submake (the command it prints) that would be
excellent as well.


I'll try, again later.

Re: Regression on Cygwin: Problems with parallel make in 4.4

[Ken Brown]

On 2/21/2023 9:39 AM, Paul Smith wrote:

On Tue, 2023-02-21 at 08:35 -0500, Ken Brown wrote:

make  check-TESTS
make[4]: Entering directory
'/home/kbrown/src/texlive/test.x86_64/Work/texk/kpathsea'
make[5]: Entering directory
'/home/kbrown/src/texlive/test.x86_64/Work/texk/kpathsea'


OK, I see the rule that generates the [5] recursion:

   $(TEST_SUITE_LOG): $(TEST_LOGS)
@$(am__set_TESTS_bases); \
 ...

(I think).  Hm, this rule does contain $(MAKE) so it should be
automatically marked for recursion.  Maybe something weird about
AM_MAKEFLAGS?

Sorry for the iteration but can you run the build again, this time with
the "--trace" option and also editing the above rule (at line 1755 of
kpathsea/Makefile to add "set -x" as in:

   $(TEST_SUITE_LOG): $(TEST_LOGS)
@$(set -x; am__set_TESTS_bases); \
 ...

There will likely be a lot of output.


Here it is.

KenMakefile:570: update target 'check-recursive' due to: target is .PHONY
fail=; \
if (target_option=k; case ${target_option-} in ?) ;; *) echo 
"am__make_running_with_option: internal error: invalid" "target option 
'${target_option-}' specified" >&2; exit 1;; esac; has_opt=no; 
sane_makeflags=$MAKEFLAGS; if { if test -z '0'; then false; elif test -n 
'x86_64-pc-cygwin'; then true; elif test -n '4.4.0.91' && test -n 
'/home/kbrown/src/texlive/test.x86_64/Work'; then true; else false; fi; }; then 
sane_makeflags=$MFLAGS; else case $MAKEFLAGS in *\\[\ \   ]*) bs=\\; 
sane_makeflags=`printf '%s\n' "$MAKEFLAGS" | sed "s/$bs$bs[$bs $bs   ]*//g"`;; 
esac; fi; skip_next=no; strip_trailopt () { flg=`printf '%s\n' "$flg" | sed 
"s/$1.*$//"`; }; for flg in $sane_makeflags; do test $skip_next = yes && { 
skip_next=no; continue; }; case $flg in *=*|--*) continue;; -*I) strip_trailopt 
'I'; skip_next=yes;; -*I?*) strip_trailopt 'I';; -*O) strip_trailopt 'O'; 
skip_next=yes;; -*O?*) strip_trailopt 'O';; -*l) strip_trailopt 'l'; 
skip_next=yes;; -*l?*) strip_trailopt 'l';; -[dEDm]) skip_next=yes;; -[JT]) 
skip_next=yes;; esac; case $flg in *$target_option*) has_opt=yes; break;; esac; 
done; test $has_opt = yes); then \
  failcom='fail=yes'; \
else \
  failcom='exit 1'; \
fi; \
dot_seen=no; \
target=`echo check-recursive | sed s/-recursive//`; \
case "check-recursive" in \
  distclean-* | maintainer-clean-*) list='auxdir/auxsub doc texk/kpathsea 
texk/ptexenc  libs utils texk' ;; \
  *) list='. doc texk/kpathsea texk/ptexenc  libs utils texk' ;; \
esac; \
for subdir in $list; do \
  echo "Making $target in $subdir"; \
  if test "$subdir" = "."; then \
dot_seen=yes; \
local_target="$target-am"; \
  else \
local_target="$target"; \
  fi; \
  (CDPATH="${ZSH_VERSION+.}:" && cd $subdir && make  $local_target) \
  || eval $failcom; \
done; \
if test "$dot_seen" = "no"; then \
  make  "$target-am" || exit 1; \
fi; test -z "$fail"
Making check in .
make[1]: Entering directory '/home/kbrown/src/texlive/test.x86_64/Work'
Makefile:1002: update target 'recurse' due to: target is .PHONY
fail= failcom='exit 1'; \
for f in x $MAKEFLAGS; do \
  case $f in \
*=* | --[!k]*);; \
*k*) failcom='fail=yes';; \
  esac; \
done; \
list='texk/kpathsea texk/ptexenc '; for one_dir in $list; do \
  if test ! -f $one_dir/Makefile; then \
test -d $one_dir || /usr/bin/mkdir -p $one_dir; \
cmd=`cat subsubdir-conf.cmd | sed "s,auxdir/auxsub,$one_dir,g"`; \
case " texk/kpathsea texk/ptexenc  " in \
  *" $one_dir "*) skip=;; \
  *) skip=' --disable-build';; \
esac; \
(cd $one_dir && echo "=== configuring in $one_dir (`pwd`)" && \
  if :; then echo "make: running /bin/sh $cmd$skip"; fi && \
  CONFIG_SHELL=/bin/sh && export CONFIG_SHELL && \
  eval /bin/sh $cmd$skip  || \
{ echo "=== configuring in $one_dir failed"; exit 1; }) && \
if test "x$skip" = x; then \
  echo "Making all in $one_dir" && \
  (cd $one_dir && make  all); \
fi || eval $failcom; \
  fi; \
done; test -z "$fail"
make[1]: Leaving directory '/home/kbrown/src/texlive/test.x86_64/Work'
Making check in doc
make[1]: Entering directory '/home/kbrown/src/texlive/test.x86_64/Work/doc'
make[1]: Nothing to be done for 'check'.
make[1]: Leaving directory '/home/kbrown/src/texlive/test.x86_64/Work/doc'
Making check in texk/kpathsea
make[1]: Entering directory 
&#x

Re: Regression on Cygwin: Problems with parallel make in 4.4

[Ken Brown]

On 2/21/2023 11:34 AM, Paul Smith wrote:

On Tue, 2023-02-21 at 10:36 -0500, Paul Smith wrote:

But, I think I might have found the bug.  If I'm right, it's a doozy!
But, as you mentioned it's not widespread because it only affects
pipe jobservers.  I'm glad 4.4.1 is almost ready to go.  I'll try to
write a test case for it to see if I'm right.


I was able to write a regression test that shows the bug.

Can you please apply the change below and see if it fixes the problem?
I don't think this is the final solution I'll apply for 4.4.1 but it
will let me know if I'm on the right track.


I think you're on the right track.  I got through 'make -j13 check' 
without the jobserver warning.


Ken

Re: Regression on Cygwin: Problems with parallel make in 4.4

[Ken Brown]

On 2/21/2023 1:26 PM, Paul Smith wrote:

On Tue, 2023-02-21 at 13:11 -0500, Ken Brown wrote:

I think you're on the right track.  I got through 'make -j13 check'
without the jobserver warning.


w00t!  Thanks for the help.  A full fix should be in 4.4.1 which I hope
to release this week or weekend.


Just as a little icing on the cake, I tried my cmake build of doxygen, 
and again it went through with no jobserver warnings.


Ken

Re: Regression on Cygwin: Problems with parallel make in 4.4

[Ken Brown]

On 2/21/2023 5:59 PM, Ken Brown wrote:

On 2/21/2023 1:26 PM, Paul Smith wrote:

On Tue, 2023-02-21 at 13:11 -0500, Ken Brown wrote:

I think you're on the right track.  I got through 'make -j13 check'
without the jobserver warning.


w00t!  Thanks for the help.  A full fix should be in 4.4.1 which I hope
to release this week or weekend.


Just as a little icing on the cake, I tried my cmake build of doxygen, 
and again it went through with no jobserver warnings.


Sorry, I spoke too soon.  I just went back to my original use case, in 
which I used the TeX Live "Build" script instead of directly invoking 
make, and I again got the jobserver warning.  [This is the use case in 
which I got a hang when using a fifo instead of a pipe.]


I'm attaching that script so you can see exactly how "make" is invoked 
in a subshell.  I'm also attaching my build log up to the point of the 
warning and the Makefile in the ft-build directory in which the warning 
occurred.  I've previously sent you the top-level Makefile.


Let me know if you need anything else.

Kenstarting TeX Live build at Thu Feb 23 15:43:01 EST 2023
CMDLINE "./Build "
PWD "/home/kbrown/src/texlive/test.x86_64/Work"
UNAME   "CYGWIN_NT-10.0-22621"
See also buildenv.log and buildinfo.log in 
/home/kbrown/src/texlive/test.x86_64/Work

+ eval ../configure --prefix=/home/kbrown/src/texlive/test.x86_64/inst 
--datadir=/home/kbrown/src/texlive/test.x86_64/inst --disable-missing 
--enable-largefile --disable-shared --with-xdvi-x-toolkit=xaw --enable-xindy 
--enable-shared --enable-silent-rules 
CLISP=/home/kbrown/src/clisp-texlive/clisp-2.49.92-1.x86_64/build/clisp.exe 
'LDFLAGS='\''-Wl,--no-insert-timestamp' '-Wl,--stack,0x80'\'''
../configure --prefix=/home/kbrown/src/texlive/test.x86_64/inst 
--datadir=/home/kbrown/src/texlive/test.x86_64/inst --disable-missing 
--enable-largefile --disable-shared --with-xdvi-x-toolkit=xaw --enable-xindy 
--enable-shared --enable-silent-rules 
CLISP=/home/kbrown/src/clisp-texlive/clisp-2.49.92-1.x86_64/build/clisp.exe 
LDFLAGS='-Wl,--no-insert-timestamp -Wl,--stack,0x80'
++ ../configure --prefix=/home/kbrown/src/texlive/test.x86_64/inst 
--datadir=/home/kbrown/src/texlive/test.x86_64/inst --disable-missing 
--enable-largefile --disable-shared --with-xdvi-x-toolkit=xaw --enable-xindy 
--enable-shared --enable-silent-rules 
CLISP=/home/kbrown/src/clisp-texlive/clisp-2.49.92-1.x86_64/build/clisp.exe 
'LDFLAGS=-Wl,--no-insert-timestamp -Wl,--stack,0x80'
checking build system type... x86_64-pc-cygwin
checking host system type... x86_64-pc-cygwin
checking for gcc... gcc
checking whether the C compiler works... yes
checking for C compiler default output file name... a.exe
checking for suffix of executables... .exe
checking whether we are cross compiling... no
checking for suffix of object files... o
checking whether the compiler supports GNU C... yes
checking whether gcc accepts -g... yes
checking for gcc option to enable C11 features... none needed
checking whether gcc understands -c and -o together... yes
checking for native WIN32 or MINGW32... no
configure: Assuming `--with-x'
configure: Assuming `--enable-autosp=yes'
configure: Assuming `--enable-axodraw2=yes'
configure: Assuming `--enable-devnag=yes'
configure: Assuming `--enable-lacheck=yes'
configure: Assuming `--enable-m-tx=yes'
configure: Assuming `--enable-pmx=yes'
configure: Assuming `--enable-ps2eps=yes'
configure: Assuming `--enable-t1utils=yes'
configure: Assuming `--enable-texdoctk=yes'
configure: Assuming `--enable-tpic2pdftex=yes'
configure: Assuming `--enable-vlna=yes'
configure: Assuming `--with-clisp-runtime=default'
configure: Assuming `--enable-xml2pmx=yes'
configure: Assuming `--enable-xpdfopen=yes'
configure: Assuming `--enable-web2c=yes'
configure: Assuming `--enable-afm2pl=yes'
configure: Assuming `--enable-bibtex-x=yes'
configure: Assuming `--enable-chktex=yes'
configure: Assuming `--enable-cjkutils=yes'
configure: Assuming `--enable-detex=yes'
configure: Assuming `--enable-dtl=yes'
configure: Assuming `--enable-dvi2tty=yes'
configure: Assuming `--enable-dvidvi=yes'
configure: Assuming `--enable-dviljk=yes'
configure: Assuming `--enable-dviout-util=yes'
configure: Assuming `--enable-dvipdfm-x=yes'
configure: Assuming `--enable-dvipng=yes'
configure: Assuming `--enable-dvipos=yes'
configure: Assuming `--enable-dvipsk=yes'
configure: Assuming `--enable-dvisvgm=yes'
configure: Assuming `--enable-gregorio=yes'
configure: Assuming `--enable-gsftopk=yes'
configure: Assuming `--enable-lcdf-typetools=yes'
configure: Assuming `--enable-makeindexk=yes'
configure: Assuming `--enable-makejvf=yes'
configure: Assuming `-

Re: Regression on Cygwin: Problems with parallel make in 4.4

[Ken Brown]

On 2/24/2023 9:54 AM, Paul Smith wrote:

On Thu, 2023-02-23 at 16:17 -0500, Ken Brown wrote:

I'm attaching that script so you can see exactly how "make" is
invoked in a subshell.  I'm also attaching my build log up to the
point of the warning and the Makefile in the ft-build directory in
which the warning occurred.  I've previously sent you the top-level
Makefile.

Let me know if you need anything else.


The ft-build/Makefile is not the interesting one: by the time make is
invoked here it already has the bad arguments.

The makefile we want is the one that invoked this make, with the bad
arguments; from your build:

   make[3]: Entering directory 
'/home/kbrown/src/texlive/test.x86_64/Work/libs/freetype2'
   rm -rf ft-build
   /usr/bin/mkdir -p ft-build
   cd ft-build && \
 CC='gcc' CONFIG_SITE=/dev/null CONFIG_SHELL='/bin/sh' \
 /bin/sh 
/home/kbrown/src/texlive/test.x86_64/Work/libs/freetype2/../../../libs/freetype2/freetype-src/configure
 \
--disable-shared \
   --without-bzip2 \
   --without-brotli \
   --without-harfbuzz \
   --without-png \
   --without-zlib \
   
--prefix=/home/kbrown/src/texlive/test.x86_64/Work/libs/freetype2/ft-install \
   --libdir=/home/kbrown/src/texlive/test.x86_64/Work/libs/freetype2 \
   --includedir=/home/kbrown/src/texlive/test.x86_64/Work/libs/freetype2

so, the makefile we want is the makefile in freetype2 that contains the
recipe that invokes the make command in the ft-build subdirectory.


OK, I hope I'm attaching the right one now.

I thought it might also help for me to send a complete recipe for 
reproducing the problem rather than just sending bits and pieces.  The 
recipe is adapted from


  https://www.tug.org/texlive/build.html

1. git clone https://github.com/TeX-Live/texlive-source.git

2. cd texlive-source

3. TL_MAKE_FLAGS="-j13" ./Build

The build is then done in texlive-source/Work, which also contains a log 
file "build.log".  On my system the jobserver warning appears after 
about 10 minutes and is in line 2074 of build.log.  I'm using make 
4.4.0.91 patched with commit 8b8cc3a825b0e9e176db5d90084c814e3744a300.


Ken

# Makefile.in generated by automake 1.16.5 from Makefile.am.
# Makefile.  Generated from Makefile.in by configure.

# Copyright (C) 1994-2021 Free Software Foundation, Inc.

# This Makefile.in is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it,
# with or without modifications, as long as this notice is preserved.

# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY, to the extent permitted by law; without
# even the implied warranty of MERCHANTABILITY or FITNESS FOR A
# PARTICULAR PURPOSE.



VPATH = ../../../libs/freetype2
am__is_gnu_make = { \
  if test -z '$(MAKELEVEL)'; then \
false; \
  elif test -n '$(MAKE_HOST)'; then \
true; \
  elif test -n '$(MAKE_VERSION)' && test -n '$(CURDIR)'; then \
true; \
  else \
false; \
  fi; \
}
am__make_running_with_option = \
  case $${target_option-} in \
  ?) ;; \
  *) echo "am__make_running_with_option: internal error: invalid" \
  "target option '$${target_option-}' specified" >&2; \
 exit 1;; \
  esac; \
  has_opt=no; \
  sane_makeflags=$$MAKEFLAGS; \
  if $(am__is_gnu_make); then \
sane_makeflags=$$MFLAGS; \
  else \
case $$MAKEFLAGS in \
  *\\[\ \   ]*) \
bs=\\; \
sane_makeflags=`printf '%s\n' "$$MAKEFLAGS" \
  | sed "s/$$bs$$bs[$$bs $$bs   ]*//g"`;; \
esac; \
  fi; \
  skip_next=no; \
  strip_trailopt () \
  { \
flg=`printf '%s\n' "$$flg" | sed "s/$$1.*$$//"`; \
  }; \
  for flg in $$sane_makeflags; do \
test $$skip_next = yes && { skip_next=no; continue; }; \
case $$flg in \
  *=*|--*) continue;; \
-*I) strip_trailopt 'I'; skip_next=yes;; \
  -*I?*) strip_trailopt 'I';; \
-*O) strip_trailopt 'O'; skip_next=yes;; \
  -*O?*) strip_trailopt 'O';; \
-*l) strip_trailopt 'l'; skip_next=yes;; \
  -*l?*) strip_trailopt 'l';; \
  -[dEDm]) skip_next=yes;; \
  -[JT]) skip_next=yes;; \
esac; \
case $$flg in \
  *$$target_option*) has_opt=yes; break;; \
esac; \
  done; \
  test $$has_opt = yes
am__make_dryrun = (target_option=n; $(am__make_running_with_option))
am__make_keepgoing = (target_option=k; $(am__make_running_with_option))
pkgdatadir = $(datadir)/freetype--tex-live-
pkgincludedir = $(includedir)/freetype--tex-live-
pkglibdir = $(libdir)/freetype--tex-live-
pkglibexecdir = $(libexecdir)/freetype--tex-live-
am__cd = CDPATH="$${ZSH_VERSION+.}$(PATH_SEPARATOR)" && cd
install_sh

Re: Regression on Cygwin: Problems with parallel make in 4.4

[Ken Brown]

On 2/24/2023 11:29 AM, Paul Smith wrote:

On Thu, 2023-02-23 at 16:17 -0500, Ken Brown wrote:

Sorry, I spoke too soon.  I just went back to my original use case,
in which I used the TeX Live "Build" script instead of directly
invoking make, and I again got the jobserver warning.


Thanks for the info.

TL;DR: this is not an error in make :)


That's a relief.  Thanks for the detailed explanation.

Ken

Re: Regression on Cygwin: Problems with parallel make in 4.4

[Ken Brown]

I'm returning to this thread because a surprising thing happened.  I 
decided to try to debug the fifo problem I reported at the beginning of 
the thread (a hang building TeX Live on Cygwin when the jobserver used a 
fifo).  So I installed make 4.4.1 built with fifos enabled (by setting 
CPPFLAGS=-DJOBSERVER_USE_FIFO=1).  And now I can no longer reproduce the 
hang.


The only explanation I can think of is that commit 8b8cc3a825, which was 
intended to fix a pipe problem, also fixed the fifo problem.  Is that 
plausible?  Or does that commit only affect the behavior of make in the 
pipe case?


Ken

Re: Regression on Cygwin: Problems with parallel make in 4.4

[Ken Brown]

On 3/2/2023 5:39 PM, Ken Brown wrote:
I'm returning to this thread because a surprising thing happened.  I 
decided to try to debug the fifo problem I reported at the beginning of 
the thread (a hang building TeX Live on Cygwin when the jobserver used a 
fifo).  So I installed make 4.4.1 built with fifos enabled (by setting 
CPPFLAGS=-DJOBSERVER_USE_FIFO=1).  And now I can no longer reproduce the 
hang.


Update: The hang occurred again.  It appears to be caused by an infinite 
loop starting with a call to pselect[*].  I looked briefly at the code 
that calls pselect, and I suspect that there is a timing issue.  Perhaps 
certain operations that are supposed to be atomic on Posix platforms are 
not atomic on Cygwin.  (Unfortunately, Cygwin's fifo implementation is 
extremely complicated in order to support multiple readers and writers, 
and atomicity had to be sacrificed.)


If I'm right, the solution would seem to be to disable the use of 
pselect on Cygwin when the jobserver is using a fifo.  I'll try that on 
a local build of make and see if I can still reproduce the problem.  It 
might be several weeks until I'm confident, since the hang occurs only 
sporadically and only after about 90 minutes of running the TeX Live build.


Ken

[*] I say "appears to be" because I was running an optimized build of 
make and an optimized build of the Cygwin DLL, so the gdb backtrace 
might not be reliable.

Re: Regression on Cygwin: Problems with parallel make in 4.4

[Ken Brown]

On 3/11/2023 12:51 PM, Ken Brown wrote:
Update: The hang occurred again.  It appears to be caused by an infinite 
loop starting with a call to pselect[*].


[...]

If I'm right, the solution would seem to be to disable the use of 
pselect on Cygwin when the jobserver is using a fifo.  I'll try that on 
a local build of make and see if I can still reproduce the problem.


Never mind.  My speculation is leading nowhere.  I just had 5 successful 
builds followed by a hang in the 6th attempt.  I guess I just have to 
continue using a pipe for my TeX Live builds.


Ken

Re: Regression on Cygwin: Problems with parallel make in 4.4

[Ken Brown]

On 3/13/2023 9:15 AM, Paul Smith wrote:

On Sun, 2023-03-12 at 11:25 -0400, Ken Brown wrote:

If I'm right, the solution would seem to be to disable the use of
pselect on Cygwin when the jobserver is using a fifo.  I'll try
that on a local build of make and see if I can still reproduce the
problem.


Never mind.  My speculation is leading nowhere.  I just had 5
successful builds followed by a hang in the 6th attempt.  I guess I
just have to continue using a pipe for my TeX Live builds.


The latest release (GNU Make 4.4.1) uses pipes for jobserver on both
Cygwin and Hurd, which both seem to have problems emulating POSIX-style
FIFOs.

Did that change not work for you, or did you by-hand force the
jobserver to use FIFOs for these tests?


We forced it to use FIFOs by configuring with -DJOBSERVER_USE_FIFO=1.

Ken

Re: [bug #64806] "invalid output sync mutex" on windows

[Ken Brown]

This is a long shot, but I had a problem a year ago with parallel make 
on Cygwin occasionally hanging.  The solution turned out to be to force 
make's jobserver to use pipes instead of fifos.  If you want to try 
this, pass make the option '--jobserver-style=pipe'.


Ken