Ergonomic, hermetic, user-friendly software build system for Python, Go, Java, Kotlin, Protobuf, Scala, Shell. Pants lets you fearlessly scale up your codebase!

Pants

platforms = [
        ‘manylinux1’,
        ‘macosx’,
    ],

Invalidated 1 target.
19:34:41 00:14   [complete]
               FAILURE
Exception caught: (&lt;class ‘pex.resolver.Unsatisfiable’&gt;)

Exception message: Could not satisfy all requirements for pulsar-client==1.19.dev2:
    pulsar-client==1.19.dev2

Yeah, you probably want the long for linux regardless since you do not support 32 bit

No cached artifacts for 1 target.
                   Invalidated 1 target.
19:36:16 00:16   [complete]
               FAILURE
Exception caught: (&lt;type ‘exceptions.ValueError’&gt;)

Exception message: invalid macosx tag: macosx

And then on OSX if it still fails, it may be the case that you want to generate the wheels without the macosx version in the name

This though will be most quickly answered by <@U054MQUPL> when he's in tomorrow

I need to check - let me do that and get back to you tomorrow

Cool - I won't be working tomorrow, but I think someone can help you

I have a general question: is there any reason why pants does not use the CRC inside the jar files? My understanding is that pants computes a checksum for every jar.

<@U48T64V7V> : we don't fingerprint 3rdparty jars as they come in

we should. but the fetching is outsourced to ivy

<@U06A03HV1> but you don't need to fingerprint them to get the CRC

<https://en.m.wikipedia.org/wiki/Zip_(file_format)>

The ZIP file format already computes CRC for the central directory, the file headers and the data descriptors

<@U48T64V7V> : yea, which i'm sure ivy checks.

*in theory*, it should be safe to get those and check if something has changed, no?

<@U06A03HV1> nope, ivy to the best of my understanding hashes everything with SHA1 (or MD5 if sha1 is missing)

but doing anything stronger, ie (committing a fingerprint to the repo as bazel does) requires fully declaring transitive deps

i was touching that code last two months when i was improving ivy resolution in sbt

<@U48T64V7V> : i suppose it depends on the purpose of the check. if the goal is to both 1) ensure that the file was transferred correctly, and 2) ensure that noone tampered with the artifact, then the CRC doesn't work

md5 doesn't do much better, as it's been broken for years

in the presence of remote artifacts, cryptographic hashes do actually become useful.