I managed to track this down.
First, I found this in sage/sage-distro/meson-logs/meson-log.txt:
*Running command: /var/tmp/sage-10.9/local/bin/gap -r -q --bare
--nointeract
-c 'Display(JoinStringsWithSeparator(GAPInfo.RootPaths,";"));'--- stdout
---ESC[?1034h/var/tmp/sage-10.9/local/lib/gap/;/var/tmp/sage-10.9/local/share/gap/---
stderr ---*
So, meson is running sage's gap executable to find its root paths. But it
is getting the "set 8 bit meta" escape sequence at the beginning of the gap
output.
If I run that command in the shell I do not get the escape sequence:
sage/local/bin/gap -r -q --bare --nointeract -c
'Display(JoinStringsWithSeparator(GAPInfo.RootPaths,";"));' | xxd
00000000: 2f76 6172 2f74 6d70 2f73 6167 652d 3130 /var/tmp/sage-10
00000010: 2e39 2f6c 6f63 616c 2f6c 6962 2f67 6170 .9/local/lib/gap
00000020: 2f3b 2f76 6172 2f74 6d70 2f73 6167 652d /;/var/tmp/sage-
00000030: 3130 2e39 2f6c 6f63 616c 2f73 6861 7265 10.9/local/share
00000040: 2f67 6170 2f0a /gap/.
However, clearly there are times when the escape sequence does get added.
The culprit turns out to be readline-6.2, which is what my docker image
happens to be using. Gemini was able to find some information about this:
* Version 6.3 Fix*: In earlier versions (notably 6.0 through 6.2),
rl_initialize() would output the "Meta mode ON" sequence to stdout
even if the output was not a terminal, causing issues like
corrupted script output and erroneous characters in program pipes.
This problem did not occur with 10.8. The difference is probably that the
rl_initialize was being called at some earlier stage of the build process,
so the escape sequence was being emitted somewhere else, not in the step
where the GAP root paths are constructed.
Moral: It is always something.
- Marc
On Tuesday, March 3, 2026 at 8:25:29 AM UTC-6 Marc Culler wrote:
> I can tell you something about where it came from.
>
> When I look at sage/local/lib/python3.14/site-packages/sage/config.py I
> see:
>
>
>
>
> *# The semicolon-separated list of GAP root paths. This is the list of#
> locations that are searched for GAP packages. This is passed directly# to
> GAP via the -l flag.GAP_ROOT_PATHS =
> "ESC[?1034h/var/tmp/sage-10.9/local/lib/gap/;/var/tmp/sage-10.9/local/share/gap/".replace("${prefix}",
>
> SAGE_LOCAL)*
>
> The file config.py is generated from sage/src/sage/config.py.in, which
> contains:
>
>
>
>
> *# The semicolon-separated list of GAP root paths. This is the list of#
> locations that are searched for GAP packages. This is passed directly# to
> GAP via the -l flag.GAP_ROOT_PATHS =
> "@GAP_ROOT_PATHS@".replace("${prefix}", SAGE_LOCAL)*
>
> So it would seem that some m4 file is inserting the "enable 8 bit meta"
> escape sequence into @GAP_ROOT_PATHS@ for my enjoyment and entertainment.
>
> Also, when I remove that escape sequence from config.py, guess what?
>
> sage: gap.FreeGroup(3)
> Group( [ f1, f2, f3 ] )
>
> - Marc
>
>
>
> On Monday, March 2, 2026 at 11:35:18 PM UTC-6 [email protected] wrote:
>
>> "\033[?1034h/var/tmp/sage-10.9/local/lib/gap/" is certainly looking bad,
>> and probably is the root of the problem.
>>
>> No idea where it came from.
>>
>>
>>
>> On Mon, Mar 2, 2026 at 9:52 PM Marc Culler <[email protected]> wrote:
>>
>>> I deleted .sage/gap and tried again. No change.
>>>
>>> I also tried
>>> sage: libgap.eval('F:= FreeGroup(3)')
>>> That produces the same GAPError, but there was some other stuff printed
>>> before the traceback which might provide a clue:
>>>
>>> #I gapdoc package is not available. Check that the name is correct
>>> #I and it is present in one of the GAP root directories (see
>>> '??RootPaths')
>>> #I GAP: needed package gapdoc cannot be loaded
>>> Error, failed to load needed package `gapdoc' (version >= 1.2)
>>> Syntax warning: Unbound global variable in
>>> /var/tmp/sage-10.9/local/share/gap/l\
>>> ib/init.g:728
>>> ColorPrompt( UserPreference( "UseColorPrompt" ) );
>>> ^^^^^^^^^^^
>>> Error, SetGasmanMessageStatus: function is not yet defined
>>> Error, Variable: 'L1_IMMUTABLE_ERROR' must have a value
>>> Error, Variable: 'ColorPrompt' must have a value
>>>
>>> I did find some other suspicious stuff:
>>>
>>> sage: libgap.eval('GAPInfo.RootPaths')
>>> [ "/home/culler/.gap/", "\033[?1034h/var/tmp/sage-10.9/local/lib/gap/",
>>> "/var/tmp/sage-10.9/local/share/gap/" ]
>>>
>>> The last path contains an empty pkg directory. The real pkg directory
>>> is in $SAGE_ROOT/local/lib/gap. But I doubt that the second path would
>>> work correctly when it has all of those ANSI escape sequences in front of
>>> it. Adding a symlink /var/tmp/sage-10.9/local/share/gap/pkg ->
>>> ../../lib/gap/pkg did not help. But maybe putting ANSI escape sequences in
>>> a root path makes GAP crash.
>>>
>>> - Marc
>>>
>>>
>>> - Marc
>>>
>>> On Mon, Mar 2, 2026 at 9:13 PM Dima Pasechnik <[email protected]> wrote:
>>>
>>>> The usual suspects for strange GAP errors are stale GAP workspaces (or
>>>> related to them errors). Make sure you don't have any (typically in
>>>> ~/.sage/, although your setup might have a different location of .sage/)
>>>>
>>>>
>>>>
>>>> On March 2, 2026 4:37:34 PM CST, Marc Culler <[email protected]>
>>>> wrote:
>>>>
>>>>>
>>>>>
>>>>> On Monday, March 2, 2026 at 3:32:23 PM UTC-6 [email protected] wrote:
>>>>>
>>>>>
>>>>> That's a problem with the pexpect GAP. Does
>>>>>
>>>>> sage: libgap.FreeGroup(3)
>>>>>
>>>>> work?
>>>>>
>>>>>
>>>>> No. I don't get a RuntimeError (which I think is produced by pexpect)
>>>>> but I get:
>>>>>
>>>>> GAPError: Error, FreeGroup: function is not yet defined
>>>>>
>>>>> Does
>>>>>
>>>>> ./sage -t src/sage/features/gap.py
>>>>>
>>>>> pass?
>>>>>
>>>>>
>>>>> Yes. I get:
>>>>>
>>>>> Doctesting 1 file.
>>>>> sage/src/bin/sage -t --warn-long 5.0
>>>>> --random-seed=206062923921806203470542245493080431476
>>>>> sage/src/sage/features/gap.py
>>>>> [5 tests, 0.01s wall]
>>>>> ----------------------------------------------------------------------
>>>>> All tests passed!
>>>>> ----------------------------------------------------------------------
>>>>> Total time for all tests: 0.2 seconds
>>>>> cpu time: 0.0 seconds
>>>>> cumulative wall time: 0.0 seconds
>>>>> Features detected for doctesting:
>>>>>
>>>>> - Marc
>>>>>
>>>>> --
>>>> You received this message because you are subscribed to a topic in the
>>>> Google Groups "sage-devel" group.
>>>> To unsubscribe from this topic, visit
>>>> https://groups.google.com/d/topic/sage-devel/dVVleHx0xP0/unsubscribe.
>>>> To unsubscribe from this group and all its topics, send an email to
>>>> [email protected].
>>>> To view this discussion visit
>>>> https://groups.google.com/d/msgid/sage-devel/B04DFD67-0325-4949-A30C-4D2A1AD9F7EE%40gmail.com
>>>>
>>>> <https://groups.google.com/d/msgid/sage-devel/B04DFD67-0325-4949-A30C-4D2A1AD9F7EE%40gmail.com?utm_medium=email&utm_source=footer>
>>>> .
>>>>
>>> --
>>>
>> You received this message because you are subscribed to the Google Groups
>>> "sage-devel" group.
>>>
>> To unsubscribe from this group and stop receiving emails from it, send an
>>> email to [email protected].
>>> To view this discussion visit
>>> https://groups.google.com/d/msgid/sage-devel/CALcZXRHNSvqL8Vi8v3fQLoOroBbi%2BXB%2B%2BSMgV8doMzU7DhLUeQ%40mail.gmail.com
>>>
>>> <https://groups.google.com/d/msgid/sage-devel/CALcZXRHNSvqL8Vi8v3fQLoOroBbi%2BXB%2B%2BSMgV8doMzU7DhLUeQ%40mail.gmail.com?utm_medium=email&utm_source=footer>
>>> .
>>>
>>
--
You received this message because you are subscribed to the Google Groups
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion visit
https://groups.google.com/d/msgid/sage-devel/f7733173-3d56-4142-af48-a12efed9b910n%40googlegroups.com.
