Don't ignore failed pre-install/post-install scripts/lua files in pkg-register(8) (#2073)

[michael-o]

When a package is registered through pkg-register(8) the return codes of pre-install and post-install scripts and lua files are ignored. Now they are respected as with pkg-add(8).

This fixes #2073

[igalic]

the title of the pull request / commit message should be what is changed, not what is wrong.

it's a commit/pull request, not an issue report

[michael-o]

the title of the pull request / commit message should be what is changed, not what is wrong.

it's a commit/pull request, not an issue report

I agree, will change

[bapt]

this is not failing on purpose for now, there are many things in the port tree that should be fixed before we do get to a point, that we stop ignore script failures.

[michael-o]

this is not failing on purpose for now, there are many things in the port tree that should be fixed before we do get to a point, that we stop ignore script failures.

Interesting..can your shortly elaborate?

Maybe those spots require comments in code depicting that this is on purpose?

[bapt]

I need to find time, to write down all of it yes, I will to make that happen in the next couple of weeks.

[michael-o]

I need to find time, to write down all of it yes, I will to make that happen in the next couple of weeks.

Merci!

[michael-o]

I need to find time, to write down all of it yes, I will to make that happen in the next couple of weeks.

Did you find the time to work on this/document it?

[bapt]

after checking the current status of the ports tree, most of the issues are gone. so it might be safe now to die on script failure.
One important left issue is the fact that some package are doing some sanity checking in pre/post scripts, this should be moved to a pre transaction sanity script. As soon as we have such feature we can merge this PR and actually fail in scripts.

[michael-o]

after checking the current status of the ports tree, most of the issues are gone. so it might be safe now to die on script failure. One important left issue is the fact that some package are doing some sanity checking in pre/post scripts, this should be moved to a pre transaction sanity script. As soon as we have such feature we can merge this PR and actually fail in scripts.

But wouldn't the outcome be the same? Failure is failure, though just better separated?

[bapt]

the outcone would be slightly different, if you die in a middle of a transaction do end up into an unstable state: half upgraded, some things might get broken more deeply than just the package that failed to installed) for example a library has been bumped all the dependent package are not update yet, but the pre install of one of them (and worst case scénario the first in the transaction) dies, ten the whole transaction is aborted, rolling back the entire transaction is impossible, the user ends up with a broken system.

With pre transaction checking, then we die early enough, an example of such scripts are in the linux packages.

[michael-o]

This explanation makes absolutely sense!