labath wrote:
> > That said, using a package designed (AIUI) for python versions for parsing
> > versions of gcc/clang does strike me as somewhat... unusual, even if it
> > does work, so _**maybe**_ there is case to be made for writing something
> > custom (I'm not sure if we really need anything more elaborate than
> > `tuple([int(part) for part in version.split(".")])`)
>
> Agreed. The first time I took a stab at this, that was the first thing I
> tried, but I quickly discovered that we had at least one tool (I think it was
> Python itself?) that contained alphabetical characters (something like
> `1.2.3rc` or `1.2.3.dev`) and I didn't feel like dealing with all the
> possible edge cases.
Ah yes, I guess *one* of the versions we are parsing is the version of python
itself..
> We have other packages the test suite relies on (pexpect, psutil, etc) so it
> seemed reasonable, but if folks feel strongly about it we can maintain our
> own "version parsing".
While I don't think we should be adding deps willy-nilly, I don't think this
one is more problematic than others in that they are all available through
common package management systems. OTOH, it has two things going against it:
- it's a hard dep breaking the ability to run any test (whereas e.g. pexpect
would only break pexpect-based tests)
- it should be relatively easy to replace
This does not constitute an endorsement of one direction of the other, I'm just
thinking out loud.
https://github.com/llvm/llvm-project/pull/93712
_______________________________________________
lldb-commits mailing list
lldb-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
