Add port action that bumps version and checksum

[l2dy (Zero King)]

Homebrew has brew bump-formula-pr. Such functionality saves a lot of work when updating many ports. Bumping ports using PortGroup like github and bitbucket should also be supported. Existing revision should be reset or removed interactively. A suggested commit message for the update should be shown.

Use cases (pseudo commands):

• port bump [[portname | pseudo-portname | port-expressions | port-url]] bumps checksums only
• port bump [[portname | pseudo-portname | port-expressions | port-url]] @<version> bumps version and checksums
• port bump --livecheck [[portname | pseudo-portname | port-expressions | port-url]] bumps both with livecheck result if ${version} == ${livecheck.version}

[Schamschula (Marius Schamschula)]

Just be mindful that numerous ports have checksums for multiple files. I personally use a python script to automatically update checksums, but this script has a list of which Portfiles need to be updated manually.

[l2dy (Zero King)]

Checksum for multiple files shouldn't be a problem, simply find and replace every old hash with the corresponding new one should work. Some ports have version set by custom methods so the version has to be updated manually. The Portfile have to be checked afterwards anyway.

[ryandesign (Ryan Carsten Schmidt)]

I use ​my portcheckup script to do this, though I have some local changes that haven't been committed yet.

Just "bumping" checksums is seldom desired; more work is needed to handle that situation correctly; see PortfileRecipes#stealth-updates.

[l2dy (Zero King)]

In the first use case I'd bump the version manually and let port(1) update the checksum.

Take R for example, it's not viable to automate the version bump in base, but the checksum could be automatically updated.

set major 3
set minor 3
set point 3
version                     ${major}.${minor}.${point}

[raimue (Rainer Müller)]

Replying to l2dy:

[...] the checksum could be automatically updated.

Checksums should be verified with the checksums published by upstream, whenever possible. I am afraid that fully automating the process will lead to maintainers being less careful. port -v checksum already prints the checksums in a format suitable for copy&paste to the Portfile.

[l2dy (Zero King)]

Replying to raimue:

port -v checksum already prints the checksums in a format suitable for copy&paste to the Portfile.

Not always suitable, there are spacing and expansion problems (e.g. ${certdata_distfile} being expanded).

I am afraid that fully automating the process will lead to maintainers being less careful.

And I forgot to remove revision lines several times which could be avoided with automation.

[danielluke (Daniel J. Luke)]

I like this idea. I think the tradeoff for making it easier for maintainers to version-bump a port is worth the potential of maintainers misuse (I doubt that the current behavior actually results in careless maintainers being more careful).

[ryandesign (Ryan Carsten Schmidt)]

Replying to l2dy:

Take R for example, it's not viable to automate the version bump in base, but the checksum could be automatically updated.

set major 3
set minor 3
set point 3
version                     ${major}.${minor}.${point}

Aberrations like that should be fixed (e.g. ​https://github.com/macports/macports-ports/pull/1371).

[l2dy (Zero King)]

I think a consistent way for port maintainers to warn others that updating this port requires special handling is needed if we add this target.

[pmetzger (Perry E. Metzger)]

Zero King: I think we underuse tests, fwiw. If you have a "port test" available, you don't have to wonder if your bump broke a port.

[mojca (Mojca Miklavec)]

See ​https://github.com/macports/macports-base/pull/120

[ryandesign (Ryan Carsten Schmidt)]

Has duplicate #60539.

[l2dy (Zero King)]

Replying to mojca:

See ​https://github.com/macports/macports-base/pull/120

Thanks. This feature has been implemented and released, so I'm going to close the ticket.

P.S. I don't like it fetching distfiles from our mirrors, because distfiles for the new version are likely not there.