I thought it was worth adding since the issue was still open in pwntools-tutorial and python 32 bit installations are still available for Windows. However, this issue in particular had been stale for quite some time (~3 years).

I consider the issue with 32 bits only applicable to Python 2.x. (Except for the tests, which could be made more portable maybe.) Also, I would prefer a warning (maybe with an interactive question asked) than a big error impossible to bypass.

What kind of logic flow did you have in mind? Based off the issue I thought it was worth stopping early since it is kind of an irrecoverable issue.

I mean that pwntools is useful on 32-bit arches as well, but some features of it may not work under python 2. Here is all I know:

• The only officially supported host platform for now is latest Ubuntu LTS on amd64 (read: this is a reference platform).
• Some of the doctests currently only work on Ubuntu LTS on amd64, because they must assume something.
• I personally tested pwntools on both arm64 (py2, py3) and i386 (py3) to work in every case I could possibly imagine.
• Historically, pwntools is known to have had broken stuff on 32-bit builds of Python 2 (one thing still broken is MIPS shellcraft). This is caused by Python 2 integers being limited to platform-dependent value ranges, and us using hex(n) or repr(n) instead of "%#x" % n or "{:#x}".format(n) in some places. Python 3 does not have such a problem, because it simply treats everything similar to longs from py2 (bignums) and never adds any L suffix.

I think I will fix all of 32-bit problems shortly, and then we can get rid of the warning, which is worded too strongly in the first place.

And I really do not like some commercial software going to great lengths trying to make 'unsupported' mean 'impossible', so I do not want pwntools to be like that. This is just being mean to all well-aware users. Yes, yes, I know advanced users can just patch the check away, but still.

We could limit the check to python 2 only. But I don't 100% understand what's broken exactly.

The other message doesn't prevent you from using pwntools on 32bit and only warns you, that there might be problems.

We could only log a warning during pip install, but I don't know if it's a good idea to stall on an input() call or similar, since it could break scripted installs.

This isn't your fault @fercevik729. You just brought up the issue again. Documenting possible incompatibility with 32bit Python 2 in the install instructions could fix the original issue in the pwntools-tutorial repo as well, without touching the setup.py?

I see. In that case should I modify the installation instructions in the README and revert my changes?