sagemathgh-36640: Use GNU make tokenpool protocol to manage paralleli…

…sm of doctesting

    
<!-- ^^^^^
Please provide a concise, informative and self-explanatory title.
Don't put issue numbers in there, do this in the PR body below.
For example, instead of "Fixes sagemath#1234" use "Introduce new method to
calculate 1+1"
-->
<!-- Describe your changes here in detail -->

This is useful when running our doctester in parallel with other build
steps, or several doctesters in parallel, as happens for example in
`make SAGE_CHECK=yes pypi-wheels`, and more of that after sagemath#35095.

To test:
```
MAKE="make -j14" make SAGE_NUM_THREADS=100 DEBUG_JOBCLIENT=1 ptest
```
This will make the doctester attempt to use 100 workers, but it will
only get tokens for 14 workers from `make`.
`DEBUG_JOBCLIENT=1` shows what's happening.

Upstream PR:
- milahu/gnumake-tokenpool#3 (merged)

<!-- Why is this change required? What problem does it solve? -->
<!-- If this PR resolves an open issue, please link to it here. For
example "Fixes sagemath#12345". -->
- Resolves sagemath#30369
<!-- If your change requires a documentation PR, please link it
appropriately. -->

### 📝 Checklist

<!-- Put an `x` in all the boxes that apply. -->
<!-- If your change requires a documentation PR, please link it
appropriately -->
<!-- If you're unsure about any of these, don't hesitate to ask. We're
here to help! -->
<!-- Feel free to remove irrelevant items. -->

- [x] The title is concise, informative, and self-explanatory.
- [ ] The description explains in detail what this PR is about.
- [x] I have linked a relevant issue or discussion.
- [ ] I have created tests covering the changes.
- [ ] I have updated the documentation accordingly.

### ⌛ Dependencies

<!-- List all open PRs that this PR logically depends on
- sagemath#12345: short description why this is a dependency
- sagemath#34567: ...
-->

<!-- If you're unsure about any of these, don't hesitate to ask. We're
here to help! -->
    
URL: sagemath#36640
Reported by: Matthias Köppe
Reviewer(s): Michael Orlitzky

build/pkgs/gnumake_tokenpool/SPKG.rst

@@ -0,0 +1,21 @@
+gnumake_tokenpool: Jobclient and jobserver for the GNU make tokenpool protocol
+==============================================================================
+
+Description
+-----------
+
+The project has implementations in multiple languages.
+
+We only install the implementation in Python.
+
+
+License
+-------
+
+MIT
+
+
+Upstream Contact
+----------------
+
+- https://github.com/milahu/gnumake-tokenpool (upstream)

build/pkgs/gnumake_tokenpool/checksums.ini

@@ -0,0 +1,5 @@
+tarball=gnumake_tokenpool-VERSION-py3-none-any.whl
+sha1=a060f03e0306a85bc1a91a450e457be83ed371e9
+md5=834ccc4d6d52741c5eabac1bdb8f39b2
+cksum=1679797266
+upstream_url=https://pypi.io/packages/py3/g/gnumake_tokenpool/gnumake_tokenpool-VERSION-py3-none-any.whl

build/pkgs/gnumake_tokenpool/dependencies

@@ -0,0 +1 @@
+ | $(PYTHON) $(PYTHON_TOOLCHAIN)

build/pkgs/gnumake_tokenpool/install-requires.txt

@@ -0,0 +1 @@
+gnumake-tokenpool

build/pkgs/gnumake_tokenpool/package-version.txt

@@ -0,0 +1 @@
+0.0.3

build/pkgs/gnumake_tokenpool/type

@@ -0,0 +1 @@
+standard

pkgs/sagemath-categories/tox.ini

@@ -26,6 +26,7 @@ passenv =
     # Parallel build
                              SAGE_NUM_THREADS
                              SAGE_NUM_THREADS_PARALLEL
+                             MAKEFLAGS
     # SAGE_VENV only for referring to the basepython or finding the wheels
     sagepython, sagewheels:  SAGE_VENV
     # Location of the wheels

pkgs/sagemath-environment/tox.ini

@@ -26,6 +26,7 @@ passenv =
     # Parallel build
                              SAGE_NUM_THREADS
                              SAGE_NUM_THREADS_PARALLEL
+                             MAKEFLAGS
     # SAGE_VENV only for referring to the basepython or finding the wheels
     sagepython, sagewheels:  SAGE_VENV
     # Location of the wheels

pkgs/sagemath-objects/tox.ini

@@ -26,6 +26,7 @@ passenv =
     # Parallel build
                              SAGE_NUM_THREADS
                              SAGE_NUM_THREADS_PARALLEL
+                             MAKEFLAGS
     # SAGE_VENV only for referring to the basepython or finding the wheels
     sagepython, sagewheels:  SAGE_VENV
     # Location of the wheels

pkgs/sagemath-repl/tox.ini

@@ -26,6 +26,7 @@ passenv =
     # Parallel build
                              SAGE_NUM_THREADS
                              SAGE_NUM_THREADS_PARALLEL
+                             MAKEFLAGS
     # SAGE_VENV only for referring to the basepython or finding the wheels
     sagepython, sagewheels:  SAGE_VENV
     # Location of the wheels

pkgs/sagemath-standard/tox.ini

@@ -75,6 +75,7 @@ passenv =
     # Parallel build
                              SAGE_NUM_THREADS
                              SAGE_NUM_THREADS_PARALLEL
+                             MAKEFLAGS
     # SAGE_VENV only for referring to the basepython or finding the wheels
     sagepython, sagewheels:  SAGE_VENV
     # Location of the wheels

src/bin/sage-runtests

@@ -26,7 +26,8 @@ if __name__ == "__main__":
                                                  ".py, .pyx, .pxd, .pxi, .sage, .spyx, .tex, .rst.")
     parser.add_argument("-p", "--nthreads", dest="nthreads",
                         type=int, nargs='?', const=0, default=1, metavar="N",
-                        help="tests in parallel using N threads with 0 interpreted as max(2, min(8, cpu_count()))")
+                        help="test in parallel using N threads, with 0 interpreted as max(2, min(8, cpu_count())); "
+                        "when run under the control of the GNU make jobserver (make -j), request as most N job slots")
     parser.add_argument("-T", "--timeout", type=int, default=-1, help="timeout (in seconds) for doctesting one file, 0 for no timeout")
     what = parser.add_mutually_exclusive_group()
     what.add_argument("-a", "--all", action="store_true", default=False, help="test all files in the Sage library")

src/doc/en/developer/doctesting.rst

@@ -512,7 +512,10 @@ doctests.  This determines the number of threads by reading the
 environment variable :envvar:`MAKE`: if it is set to ``make -j12``, then
 use 12 threads.  If :envvar:`MAKE` is not set, then by default it uses
 the number of CPU cores (as determined by the Python function
-``multiprocessing.cpu_count()``) with a minimum of 2 and a maximum of 8.
+:func:`multiprocessing.cpu_count`) with a minimum of 2 and a maximum of 8.
+(When this runs under the control of the `GNU make jobserver
+<https://www.gnu.org/software/make/manual/make.html#Parallel>`_, then Sage
+will request as most this number of job slots.)
 
 In any case, this will test the Sage library with multiple threads::
 

src/doc/en/installation/source.rst

@@ -782,14 +782,21 @@ Here are some of the more commonly used variables affecting the build process:
       Some users on single-core macOS machines have reported problems when
       building Sage with ``MAKE='make -jNUM'`` with ``NUM`` greater than one.
 
-- :envvar:`SAGE_NUM_THREADS` - if set to a number, then when building the
-  documentation, parallel doctesting, or running ``sage -b``, use this many
-  threads.
+- :envvar:`SAGE_NUM_THREADS` - if set to a number, then when rebuilding with
+  ``sage -b`` or parallel doctesting with ``sage -t -p 0``, use at most this
+  many threads.
+
   If this is not set, then determine the number of threads using the value of
   the :envvar:`MAKE` (see above) or :envvar:`MAKEFLAGS` environment variables.
-  If none of these specifies a number of jobs, use one thread (except for
-  parallel testing: there we use a default of the number of CPU cores, with a
-  maximum of 8 and a minimum of 2).
+  If none of these specifies a number of jobs,
+
+  - ``sage -b`` only uses one thread
+
+  - ``sage -t -p 0`` uses a default of the number of CPU cores, with a
+    maximum of 8 and a minimum of 2.
+
+  When ``sage -t -p`` runs under the control of the GNU ``make``
+  jobserver, then Sage will request as most this number of job slots.
 
 - :envvar:`V` - if set to ``0``, silence the build.  Instead of
   showing a detailed compilation log, only one line of output is shown

src/sage/doctest/forker.py

@@ -1803,6 +1803,16 @@ def parallel_dispatch(self):
         """
         opt = self.controller.options
 
+        job_client = None
+        try:
+            from gnumake_tokenpool import JobClient, NoJobServer
+            try:
+                job_client = JobClient()
+            except NoJobServer:
+                pass
+        except ImportError:
+            pass
+
         source_iter = iter(self.controller.sources)
 
         # If timeout was 0, simply set a very long time
@@ -1925,6 +1935,9 @@ def sel_exit():
                             w.copied_pid = w.pid
                             w.close()
                             finished.append(w)
+                            if job_client:
+                                job_client.release()
+
                     workers = new_workers
 
                     # Similarly, process finished workers.
@@ -1959,11 +1972,14 @@ def sel_exit():
                         break
 
                     # Start new workers if possible
-                    while source_iter is not None and len(workers) < opt.nthreads:
+                    while (source_iter is not None and len(workers) < opt.nthreads
+                           and (not job_client or job_client.acquire())):
                         try:
                             source = next(source_iter)
                         except StopIteration:
                             source_iter = None
+                            if job_client:
+                                job_client.release()
                         else:
                             # Start a new worker.
                             import copy
@@ -2040,6 +2056,8 @@ def sel_exit():
                         sleep(die_timeout)
                         for w in workers:
                             w.kill()
+                            if job_client:
+                                job_client.release()
                     finally:
                         os._exit(0)