prompt_toolkit 3.0.25+ breaks Ctrl-C

[tornaria]

For the record, Ctrl-C is broken starting with version 3.0.25 of prompt_toolkit (current version 3.0.28 still broken, sage-the-distribution bundles 3.0.22 so it's ok).

$ xbps-query -s python3-prompt_toolkit
[*] python3-prompt_toolkit-3.0.28_1 Python3 library for building powerful interactive command lines
$ sage
┌────────────────────────────────────────────────────────────────────┐
│ SageMath version 9.5, Release Date: 2022-01-30                     │
│ Using Python 3.10.2. Type "help()" for help.                       │
└────────────────────────────────────────────────────────────────────┘
sage: factor(54853908712446157179434453567343931596301333824416758692208077566613224454501131209)
^C^C^C^C

If one waits the half hour or so that it takes to factor the number above, then the ^C gets through, but waiting kind of defeats the purpose of hitting ^C. This is not specific to factoring -- anything that depends on cysignals for Ctrl-C working is broken.

Upstream issue: prompt-toolkit/python-prompt-toolkit#1576

In the GH issue there are two different fixes suggested (with patches), IMHO the second one is better for sage (the downside is that introduces a dependency on cysignals, but sage already depends on cysignals).

No action is needed now, but one of the patches will be necessary if/when prompt_toolkit is updated, or for distro packages on systems with a recent version of prompt_toolkit (void linux issue and discussion in void-linux/void-packages#35730)

Upstream: Reported upstream. No feedback yet.

CC: @antonio-rojas @orlitzky @jhpalmieri

Component: packages: standard

Keywords: prompt_toolkit

Author: Matthias Koeppe

Branch/Commit: bd1bc04

Reviewer: John Palmieri

Issue created by migration from https://trac.sagemath.org/ticket/33428

[mkoeppe]

comment:2

Let's put in an upper version bound in build/pkgs/prompt_toolkit/install-requires.txt to document this

[mkoeppe]

Branch: u/mkoeppe/prompt_toolkit_3_0_25__breaks_ctrl_c

[mkoeppe]

Commit: bd1bc04

[mkoeppe]

New commits:

bd1bc04

build/pkgs/prompt_toolkit/install-requires.txt: Reject versions 3.0.25+

[mkoeppe]

Author: Matthias Koeppe

[mkoeppe]

comment:5

Upstream has merged prompt-toolkit/python-prompt-toolkit#1582 - do we know if it fixes the issue?

[tornaria]

comment:6

Replying to @mkoeppe:

Upstream has merged prompt-toolkit/python-prompt-toolkit#1582 - do we know if it fixes the issue?

I don't think so, we are patching 3.0.29 in void linux, see here:
https://github.com/void-linux/void-packages/blob/master/srcpkgs/python3-prompt_toolkit/patches/dont-handle-sigint.patch

[jhpalmieri]

comment:8

Confirmed: I just downloaded and installed 3.0.30 (OS X) and it does not seem to have fixed the interrupt problems.

[jhpalmieri]

Reviewer: John Palmieri

[jhpalmieri]

comment:9

Is this going to cause problems with the IPython upgrade (#33530)? That branch uses version 3.0.29.

[mkoeppe]

comment:10

Thanks for catching this.

[vbraun]

comment:11

Merge failure on top of:

1c98f0a Trac #31563: Upgrade giac to 1.9.0-15

83a3df1 Trac #34131: some doc polishing in symbolic

dee5229 Trac #34130: some doc polishing in groups

fd80fa1 Trac #34126: some doc polishing in categories

718d356 Trac #34125: some doc polishing in modular

49390d8 Trac #34124: some doc polishing in combinat

af94c25 Trac #34096: Fix typo: enviroment -> environment

d04c189 Trac #34050: add some space around == in paths and rational

1634bb6 Trac #34047: Typo in _repr_Expression in symbolic/expression.pyx

dc8086f Trac #34154: Fix repeated word typos

17b1c17 Trac #34148: fix and activate pycodestyle E306

785bdbd Trac #34136: modernize super() in rings (part one)

9882001 Trac #34129: Dummy packages should write proper log files

dbc6b9e Trac #34109: pep8 for sagedoc.py

5e6a331 Trac #34103: randomize infinite recursion

17ba76f Trac #34093: Wrong evaluation of elements of CBF[x] on polynomials from other rings

4657052 Trac #34122: Bug in is_planar() method for directed graphs

c9cfb53 Trac #34087: inverse reciprocal trig functions not (back)translated to/from Mathematica

1d6d055 Trac #34076: pycodestyle cleanup in src/sage/graphs/genus.pyx

043d862 Trac #34071: pycodestyle cleanup in asteroidal_triples.pyx, chrompoly.pyx, cliquer.pyx, convexity_properties.pyx

b9b25dc Trac #34070: pycodestyle cleanup in src/sage/graphs/centrality.pyx

4ef2c65 Trac #34069: pycodestyle cleanup in src/sage/graphs/comparability.pyx

5b11467 Trac #34066: Tachyon help message

798adaa Trac #34065: pycodestyle cleanup in src/sage/graphs/bliss.pyx

ce62be2 Trac #34063: pycodestyle cleanup in src/sage/graphs/base/

a0eadb3 Trac #34059: Fix trivial case in conversion of list to number field element

fea0ac5 Trac #34057: changes about avoiding double dieses

9eefd5c Trac #34145: modernize super in geometry/

01e117d Trac #34137: modernize super in categories/

6c79334 Trac #33819: build.yml: In workflow_dispatch, make container and base tag selectable; add doc

6a64ab8 Trac #33760: do not update _pos, _pos3d, _embedding on vertex/edge deletions

dbf091d Trac #34139: fix the linter

625ac58 Updated SageMath version to 9.7.beta5

ticket milestone is not intended to be merged:

[vbraun]

Changed branch from u/mkoeppe/prompt_toolkit_3_0_25__breaks_ctrl_c to bd1bc04

[mkoeppe]

@tornaria Do you know if there's a new prompt-toolkit version that does not need to be patched?
IPython 8.11 wants a prompt-toolkit >= 3.0.30

[tornaria]

@tornaria Do you know if there's a new prompt-toolkit version that does not need to be patched? IPython 8.11 wants a prompt-toolkit >= 3.0.30

We are shipping 3.0.38 in void linux, and I'm still patching it (see https://github.com/void-linux/void-packages/blob/master/srcpkgs/python3-prompt_toolkit/patches/dont-handle-sigint.patch)

But I haven't tested it without the patch so... maybe something else changed somewhere (in ipython?). My suggestion would be to upgrade ipython to 8.11 and prompt-toolkit to 3.0.38, and if the bug is still present you can apply the patch I linked above.

[mkoeppe]

Thanks. I've just tested it; still broken (and prompt-toolkit/python-prompt-toolkit#1576 is still open)

[tornaria]

@mkoeppe Here is a possible workaround ( "if the mountain won't come to sagemath..." )

What do you think?

Note that the patch I linked above is a different workaround which is not so good (completely disabling the sigint handler in prompt_toolkit carries other problems: they added this handler in 3.0.25 for a reason, see prompt-toolkit/python-prompt-toolkit#1538).

--- a/src/sage/repl/interpreter.py
+++ b/src/sage/repl/interpreter.py
@@ -153,6 +153,8 @@ from IPython.terminal.embed import InteractiveShellEmbed
 from IPython.terminal.ipapp import TerminalIPythonApp, IPAppCrashHandler
 from IPython.core.crashhandler import CrashHandler
 
+from cysignals import pysignals
+import signal
 
 # TODO: This global variable _do_preparse should be associated with an
 # IPython InteractiveShell as opposed to a global variable in this
@@ -287,6 +289,13 @@ class SageTerminalInteractiveShell(SageShellOverride, TerminalInteractiveShell):
         backend = BackendIPythonCommandline()
         backend.get_display_manager().switch_backend(backend, shell=self)
 
+    def prompt_for_code(self):
+        _sigint = signal.getsignal(signal.SIGINT)
+        _sigint_os = pysignals.getossignal(signal.SIGINT)
+        text = TerminalInteractiveShell.prompt_for_code(self)
+        pysignals.setsignal(signal.SIGINT, _sigint, _sigint_os)
+        return text
+
 
 class SageTestShell(SageShellOverride, TerminalInteractiveShell):
     """