boltons.ecoutils 23.0.0 breaks pdb interactive prompt in pytest debug sessions

[kdeldycke]

I've got a really strange issue.

Tl;Dr: when I import from boltons.ecoutils import get_profile from boltons 23.0.0, the interactive pdb prompt from my pytest --pdb calls disappears.

Before the upgrade to 23.0.0, I had:

$ poetry run pytest --pdb
================= test session starts =================
platform darwin -- Python 3.11.2, pytest-7.2.1, pluggy-1.0.0
Using --randomly-seed=1875232546
plugins: cases-3.6.13, randomly-3.12.0, cov-4.0.0, httpserver-1.0.6
(...)
click_extra/tests/test_colorize.py 
>>>>>>> PDB set_trace (IO-capturing turned off) >>>>>>>
--Call--
> .../lib/python3.11/site-packages/click/testing.py(133)output()
-> @property
(Pdb) 

After the upgrade to 23.0.0, I get a blocking prompt:

$ poetry run pytest --pdb
================= test session starts =================
platform darwin -- Python 3.11.2, pytest-7.2.1, pluggy-1.0.0
Using --randomly-seed= 1875232546
plugins: cases-3.6.13, randomly-3.12.0, cov-4.0.0, httpserver-1.0.6
(...)
click_extra/tests/test_colorize.py 
>>>>>>> PDB set_trace (IO-capturing turned off) >>>>>>>
--Call--
> .../lib/python3.11/site-packages/click/testing.py(133)output()
-> @property

See above the blocking pdb session without any prompt. It is not responding to any pdb command and have to exits with ^D.

And in the configuration above, as soon as I remove the from boltons.ecoutils import get_profile statement from my code, while keeping the 23.0.0 version of boltons installed, then the problem disappear and I get back my interactive pdb prompt.

I confirm this issue to be completely absent from the 21.0.0 release of boltons.

For reference, the issue appeared in my project with commit: kdeldycke/click-extra@1f7c6de

As for the potential fix, there something wrong with the changes in boltons/ecoutils.py between 21.0.0 and 23.0.0: 21.0.0...23.0.0#diff-71d9829db55b3d86e1aefcc41e09ab86f3de1d42db1d9859f9976413f0315c86 . Probably with the way pprint._safe_repr is monkey-patched.

[kdeldycke]

OK, I think I found the root cause. A fix is proposed at: #335

[kdeldycke]

Forget about my patch proposed in #335. The real culprit is readline. I missed it as my code editor made some linting changes behind my back.

So, to sum it up, this patch fix the issue described here:

diff --git boltons/ecoutils.py boltons/ecoutils.py
index c5ff16a..6bf3367 100644
--- boltons/ecoutils.py
+++ boltons/ecoutils.py
@@ -180,7 +180,7 @@ HAVE_UCS4 = getattr(sys, 'maxunicode', 0) > 65536
 HAVE_READLINE = True
 
 try:
-    import readline
+    pass
 except Exception:
     HAVE_READLINE = False
 

Of course this is not an acceptable change to push as-is to upstream, as we loose the readline detection. But at least I found the culprit.

I'll try to dig for a root cause in that direction. In the mean time, here is the system I'm running:

• Python 3.11.3
• macOS Ventura 13.3.1
• readline: stable 8.2.1 (bottled) [keg-only] (installed via brew)

[mahmoud]

Veeeerry interesting. readline has always been a tricky one. I don't have any immediate workarounds that come to mind, but I'm looking forward to anything else you find.

[kdeldycke]

Might be related:

• Prevent pyreadline from bugging out in --pdb mode with capture pytest-dev/pytest#1281
• Pytest warning on readline pytest-dev/pytest#8847
• Drop readline workaround introduced in #1281 pytest-dev/pytest#8848