Bumps [docker/setup-qemu-action](https://github.com/docker/setup-qemu-action) from 1 to 2.
- [Release notes](https://github.com/docker/setup-qemu-action/releases)
- [Commits](docker/setup-qemu-action@v1...v2)

---
updated-dependencies:
- dependency-name: docker/setup-qemu-action
  dependency-type: direct:production
  update-type: version-update:semver-major
...

Signed-off-by: dependabot[bot] <support@github.com>

read_text/write_text default to the locale encoding, which may or may
not be utf8.  Fix these by making the encoding explicit or by using
bytes.

I don't think other encodings really worked previously, so there's no
deprecation period.

For `FontProperties(stretch=12345)`; before:
```
Traceback (most recent call last):
  File ".../matplotlib/font_manager.py", line 888, in set_stretch
    raise ValueError()
ValueError

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File ".../matplotlib/font_manager.py", line 695, in __init__
    self.set_stretch(stretch)
  File ".../matplotlib/font_manager.py", line 891, in set_stretch
    raise ValueError("stretch is invalid") from err
ValueError: stretch is invalid
```
after
```
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File ".../matplotlib/font_manager.py", line 695, in __init__
    self.set_stretch(stretch)
  File ".../matplotlib/font_manager.py", line 900, in set_stretch
    raise ValueError(f"{stretch=} is invalid")
ValueError: stretch=12345 is invalid
```

For tex failures (e.g., `figtext(.5, .5, "\N{snowman}", usetex=True)`),
instead of
```
Traceback (most recent call last):
  File ".../matplotlib/texmanager.py", line 253, in _run_checked_subprocess
    report = subprocess.check_output(
  File "/usr/lib/python3.10/subprocess.py", line 420, in check_output
    return run(*popenargs, stdout=PIPE, timeout=timeout, check=True,
  File "/usr/lib/python3.10/subprocess.py", line 524, in run
    raise CalledProcessError(retcode, process.args,
subprocess.CalledProcessError: Command '['latex', '-interaction=nonstopmode', '--halt-on-error', '../71cab2b5aca12ed5cad4a481b3b00e42.tex']' returned non-zero exit status 1.

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
    <a long traceback>
    raise RuntimeError(
RuntimeError: latex was not able to process the following string:
b'\\u2603'

Here is the full report generated by latex:
This is pdfTeX, Version 3.141592653-2.6-1.40.23 (TeX Live 2021) (preloaded format=latex)
<the tex output>
```

instead report the failure in a single block, as
```
Traceback (most recent call last):
    <a long traceback>
    raise RuntimeError(
RuntimeError: latex was not able to process the following string:
b'\\u2603'

Here is the full command invocation and its output:

latex -interaction=nonstopmode --halt-on-error ../71cab2b5aca12ed5cad4a481b3b00e42.tex

This is pdfTeX, Version 3.141592653-2.6-1.40.23 (TeX Live 2021) (preloaded format=latex)
<the tex output>
```
(the exception chaining is not particularly informative, and we can move
the only relevant info -- the command we tried to run -- to the second
exception).

This fixes the following error in mingw gcc toolchain. Also clang
also have same error.

src/_tkagg.cpp:274:26: note:   crosses initialization of 'bool tk_ok'
  274 |     bool tcl_ok = false, tk_ok = false;
      |                          ^~~~~
src/_tkagg.cpp:274:10: note:   crosses initialization of 'bool tcl_ok'
  274 |     bool tcl_ok = false, tk_ok = false;
      |          ^~~~~~

According to C++ standard (6.7.3):
It is possible to transfer into a block, but not in a way that bypasses
declarations with initialization. A program that jumps from a point where
a variable with automatic storage duration is not in scope to a point
where it is in scope is ill-formed unless the variable has scalar type...

This adds the rcParams savefig.directory option into the macosx
backend for the savefig dialog window.

fill_between isn't the right thing to use here. It might do the same
thing (as long as the plot isn't interacted with / zoomed out), but it
feels real clunky and should not be encouraged to use here in a user
facing example. I only encountered this because this plot was showing up
as a usage example on the plt.fill_between() API docs (older version)
and I found this real odd.

Arrays sometimes don't have all the methods arrays should have, so
add another check here.  Plot requires both ndim and shape and this
will extract the numpy array if x does not have those attributes.
Otherwise leave the object alone, because unit support (currently only
in plot) requires the object to retain the unit info.

When an overunder symbol has a subscript, it goes below the symbol, and the baseline of the Vlist thus created is actually the baseline of the lowest item in the Vlist. Hence, it must be moved only if there is a subscript, and not otherwise.

Adds compress=True to compressed layout engine.  Works for compact
axes grids.

Closes matplotlib#23239.

Closes matplotlib#23114, where somebody has installed matplotlib into another
git repo.

…y code

The *math* parameter is passed through many layers of the call stack
but is ultimately only used for a single purpose: deciding whether to
replace the ASCII hyphen by a (longer) unicode minus.  Instead of doing
that, just do the substitution at the parsing stage.  In particular,
this fixes problematic unicode minus support with the "cm" fontset.

This patch also reverts a significant part of 52003e4, as LogFormatters
no longer need to pass unicode minuses in mathtext -- everything gets
converted by mathtext.  Likewise, this change also invalidates the
test_log_scales baseline image (old, buggy wrt. unicode minus); replace
it by a test that the drawn ticks are as expected (which was the intent
in 90c1aa3).

AFAICT, (incorrect) support for the double-backslashed version was
accidentally introduced in 027dd2c.

This also ensures consistency in the names in the error strings; see
e.g. changes in test_mathtext (the contents of `\overline` are indeed a
"body", not a "value").

Renaming "simple_group" to "optional_group" is semantically more
meaningful.