Hook : Sort all hooks following an human listing

[camlafit]

Gitea execute hook following a glob listing, in some case this list could be not adapted. Mainly if a hook required to be execute at specific step.

Prefer a $(ls -1 |sort -h) looks useful.

Then with this list : 01-erere ezaezae 02-erezrze 1-zerezr 10-erezrze 2-erezrze 99-ererer

we obtain actually :
01-erere
02-erezrze
10-erezr
1-zerezr
2-erezrze
99-ererer
ezaezae

As we can see 10 is previous 1

Should be better to get this result :
ezaezae
01-erere
1-zerezr
02-erezrze
2-erezrze
10-erezrze
99-ererer

[bkcsoft]

This is standard .d behaviour. I don't see how this would be useful TBH :/

Edit: If you need them to be executed in a specific order just prefix them with NN- (where N is a digit)

[camlafit]

Hi I've forgotten to provide a sample, I've an use case with subgit (hook system to translate git<->svn)

…

I would also like to point to one thing: in pre-receive.d/ it would be better if a symlink to SubGit hook would be called the last, so maybe it's worth renaming the hooks to start with numbers if Gitea respects names order: pre-receive.d/ 00-hook1 01-hook2 ... 99-symlink-to-subgit-hook This is because SubGit could translate something to SVN but if there were a hook running after SubGit hook, this hook could reject the whole push. As result there would be a corresponding commit in SVN while "git push" command would fail (and then SubGit will translate this new SVN revision back to Git and the result will be as if that later hook didn't fail at all). It's not a big problem but just a strange behaviour. E.g.this would be especially strange if the hook running after SubGit hook would control access to this or that branch, because its result would be completely ignored. We had this problem with Gitolite and for it we recommended to make sure Git hooks are called the last.

[camlafit]

As said on mine examples, NN prefix is not relevant we can have 10 before 1
Why ls with sort could be solve this behavior

[bkcsoft]

NN prefix works just fine, 01_ is listed before 10_ 🙂

[camlafit]

but could be better to have a predictable sort without considerate separator :)
And if I've understood related patch should be small

[bkcsoft]

But it is predictable. This is literaly how every single conf.d implementation works 😕

[camlafit]

Hi Ok as won't be changed, maybe should be better to provide a documentation about hooks behavior and naming

[bkcsoft]

That does however make sense :)

[warnat]

can this be closed now, to get gitea forward, please?