Index | Thread | Search

From:
Crystal Kolipe <kolipe.c@exoticsilicon.com>
Subject:
Re: pax(1): new -o keyword framework, listopt, global exthdrs, and stricter invalid-path handling
To:
David Uhden Collado <daviduhden@gmail.com>
Cc:
tech@openbsd.org
Date:
Sun, 12 Oct 2025 11:00:13 +0100

Download raw body.

Thread
On Sun, Oct 12, 2025 at 10:28:35AM +0200, David Uhden Collado wrote:
> I have attached a patch to this email that modernizes pax/tar to match
> POSIX.1-2024 and adds a comprehensive -o keyword framework.

[ ... ]

For such extensive modifications to a critical program in base, you really
want to split the diff up in to smaller chunks.  An exception to this might be
if you're just showing 'work in progress' to get some early feedback, but this
doesn't seem to be the case here.

Also, there are regression tests for pax in src/regress/bin/pax.  It's useful
to mention that your changes pass the existing regress tests, (if you read
posts to -tech over a period of time, you'll see this quite frequently).

(Once new features are accepted, it might even be worth proposing new regress
 tests relevant to them, although this is by no means essential.)

Your changes to the standards section of the manual page change it to say that
pax is now compliant with the spec.  Although you have indeed have added
support for the functions that were obviously missing, have you actually
checked that the existing code is 100% compliant, and that this claim is
valid?

(Some time ago during recent changes to pax, there was a discussion about how
 to handle character encodings, my understanding is that either a decision was
 taken to specifically not follow the spec to the letter, or there was not
 full agreement between everyone about what the spec was actually demanding.)

Finally, remember that a new -release is just around the corner.  This
probably isn't a great moment to propose such a large change.