Implement RPM-free mode

[nforro]

specfile uses RPM for macro expansion and condition evaluation, however, that's not always necessary, hence the force_parse argument to the Specfile constructor that allows to use the created instance even if RPM parsing fails. But since specfile is already capable of parsing macro definitions as well as macro substitutions, we could go one step further and implement a simple, self-contained macro expansion, without using RPM.

[lachmanfrantisek]

Two main benefits:

• No need to rely on RPM.
• Be able to parse parts of specfile or specfiles that are not parsable as a whole (e.g. on a particular system).

This might be highly beneficial for Copr as well.

The solution might be in the form of a different mode people can switch to.

@praiskup will start a community discussion to gather some feedback.

[ssbarnea]

I don't know how state how important this is. With the current dependency on https://pypi.org/project/rpm/ - packit is not effectively usable and sometimes even not even installable on: macos, windows and ubuntu and we cannot even lint (validate) the config.

By default packit should not list binary dependencies directly, so it can install easily. The extra dependencies can be added to extras.

[lachmanfrantisek]

Thanks @ssbarnea for your interest. Sadly, currently, Packit is quite dependent on rpm. Both via specfile library and by direct use as well. This issue can help with specfile parsing and if you require only validation, we would need to factor that out to not require rpm.

There is also a workaround in the form of containers -- I hope adding more architecture can help here at least a bit -- here are the fresh PRs:

• feat(base): build for alternative architectures deployment#620
• ci: build containers for alternative architectures packit#2485

---

To be really honest and to better understand why it's taking us so long to tackle this issue. Our team prioritises work based on demand and there haven't been many people requesting this so far. But if we know there are enough people, we are happy to spend time on this. As usual, an external contribution can help and we are happy to guide anyone willing to help. (I know, it's not easy.)

[nforro]

Still relevant.

[praiskup]

@praiskup will start a community discussion to gather some feedback.

Uh. The norpm project progressed since 2024, and we seem to be able to parse the vast majority of specfiles with plain Python = without real RPM being installed (ignoring shell/lua macros for security reasons, but we can opt-in to parse those, too). For example, the rpmautospec project is already using norpm as opt-in.

I barely remember I promised some discussion here though. And I would hate to block any movement here :-( do we still need to discuss something with Fedora community?