Arch Linux User Repository

@m040601: I removed fish from checkdepends().

First of all, thanks for your work in this PKGBUILD. Really cool use of color and tests.

I see that the "fish dependency question" has already been raised and discussed. And just like others, I also think that,

you don't really even need fish to successfully pass the tests

I use bash, and this package was working fine in the last months without any type of unnecessary fish dependency. I really dont see why do I have to download and install fish to my system, if I'm not gonna use it. Even if I'm throwing it out away immediattely after. Be it a "check" or "make" dependency.

@yochananmarqos @devnore: For what it's worth, you don't really even need fish to successfully pass the tests.

If fish is missing, the relevant test is actually skipped. The ./test.sh script won't complain at all (unless you use --strict), and the ./build.sh script will only give a small warning about skipped tests:

Verifying scripts...
batpipe:detected_fish_shell skipped.
One or more tests were skipped.
Run ./test.sh for more detailed information.
Verified successfully.

P.S. The main reason I have it as a test in the first place is because reliably detecting the user's login shell is pretty brittle and platform-dependent, and that was something that I felt should be automatically tested by the CI.

@yochananmarqos but i still need to install fish to install. (or is there a switch in yay/paru to skip testing phase?)

@devnore: fish is only required for the tests. Notice it's in checkdepends()?

@yochananmarqos what is the reasoning for adding fish as a dependency to a set of bash-scripts?

@yochananmarqos I have orphaned the package, feel free to take it over. I only use batman now among the scripts, so I'm not the best person to maintain it.

@desbma:

There are some missing optional dependencies:

optdepends=('ripgrep: required for batgrep script'
            'entr: optional for batwatch script'
            'ncurses: optional for batdiff script'
            'prettier: various formatting for prettybat'
            'shfmt: bash formatting for prettybat')

I'm a bit late to comment here, but there's a couple things that can be done about some of the issues mentioned:

The build script does not support separating building from testing.

A little while ago, I changed how build.sh works. If you pass --no-verify as a parameter, it won't attempt to run tests on the built "bin" scripts.
For testing in a separate step, you can run ./test.sh --compiled after running the build.

This does not build without ripgrep installed.

If you only want to test specific test suites, that can be done with ./test.sh --compiled --suite batgrep. The list of valid suite names can currently be found with this command:

for suite in test/suite/*.sh; do basename "$suite" .sh; done

the authors of bat-extras explicitely recommend using the git master branch instead of a release

I'm not explicitly against using a release build, but I recommended using the git master since I haven't established a proper release cycle yet.

Git tags are now available

With the latest release, I changed the version numbers recently to be yyyy.mm.dd instead of yyyymmdd, since the previous format wasn't as clear about what it represented as I felt it should have been.

If anybody has any questions or issues about bat-extras, I would be happy to help with them. The best way is to make an issue on the GitHub repository, but I'll be trying to keep an eye out for anything that comes up here as well.

I won't add provides/conflicts for an hypothetical package that does not exist, especially when the authors of bat-extras explicitely recommend using the git master branch instead of a release.

That's logical, no worries.

I guess you have a varying standard regarding useless variables

I never said useless, I said unnecessary. There's no reason to use a custom variable when a predefined one is available.

However, it is of course permissable to prefix custom variables with an underscore as not to conflict with existing or future predefined variables. For more info, see Package etiquette.

I see nothing in the packaging guidelines about removing the '-git' suffix from the package name when installing the LICENSE files.

The name of the project is bat-extras no matter what package extension it has here. Therefore any files and folders should reflect that.

There's a reason the ${pkgname%-*} variable is used throughout the VCS PKGBUILD template instead of $pkgname.