python3: 'filter' object is not subscriptable in libs/gap/util.pyx

[fbissey]

Because sage-on-gentoo doesn't use GAP_DIR we are exposed to a code path that is not doctested and that fails with python3.

      File "sage/libs/gap/util.pyx", line 199, in sage.libs.gap.util.initialize (/dev/shm/portage/sci-mathematics/sage-9999/work/sage-9999/src-python3_6/build/cythonized/sage/libs/gap/util.c:4452)
        s = str_to_bytes(gap_root(), FS_ENCODING, "surrogateescape")
      File "sage/libs/gap/util.pyx", line 171, in sage.libs.gap.util.gap_root (/dev/shm/portage/sci-mathematics/sage-9999/work/sage-9999/src-python3_6/build/cythonized/sage/libs/gap/util.c:4232)
        gapdir = filter(lambda dir:dir.strip().startswith('GAP_DIR'), gap_sh)[0]
    TypeError: 'filter' object is not subscriptable

[fbissey]

Still have to write a doctest.

---

New commits:

​e273d65

Replace filter for compatibility with python3

[fbissey]

I am not sure how to write that doctest. To test this properly we have to start a new instance of sage with GAP_ROOT_DIR set to wrong value and then call gap_root. I cannot seem to change the value inside of sage in a way that sticks until the gap_root call.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​842c6df

Figured out a way of doctesting gap_root alternate code

[fbissey]

Figured something out.

[tscrim]

LGTM.

[embray]

I'm not wild about the fact that this will just crash with a not-so-useful exception if the file is not found, or the line being searched for in that file is not found. But that was the case before this ticket as well so it's fine. I just thought that was worth pointing out.

[fbissey]

Hopefully once the new gap package is sorted all that mess will disappear.

[gh-timokau]

sage: os.system("GAP_ROOT_DIR=/not_a_path sage -c \"sage.libs.gap.util.gap_root()\"")
The gap-4.5.5.spkg (or later) seems to be not installed!

This breaks for me because I set GAP_ROOT_DIR in sage-env which is sourced on sage startup. As far as I understand that is how sage-env and GAP_ROOT_DIR are intended to be used, so the doctest should be changed. Also I would expect this test to add significant overhead for little gain, since it has to go through the whole sage startup.

[gh-timokau]

Fix in #26758.