Add FIXUP option, for automatic package hash.

[cmangla]

Fixes #18
Fixes #35

[cmangla]

It seems FIXUP=1 needs to be set when PKG_MIRROR_HASH:=skip is set. See:
https://openwrt.org/docs/guide-developer/packages#use_source_repository

[cmangla]

@aparcar Please do "Squash and merge" if the PR is good and whenever it is convenient for you. Thanks!

[qwerttvv]

This PR worked perfect, THANKS!!!

[EkkoG]

I think when FIXUP=1, package should not compile, or there can have another option to enable fixup-only action

[aparcar]

I don't understand the motivation for this PR. Could you please explain and add it to the commit message?

[cmangla]

I don't understand the motivation for this PR. Could you please explain and add it to the commit message?

Let me run the motivation by you for your feedback, before I update the commits.

The motivation behind adding the FIXUP option is to generalize my specific motivation, which is to be able to compile a single package that sets PKG_MIRROR_HASH=skip. Others have expressed similar desires, such as in Issue #35 .

When PKG_MIRROR_HASH=skip and the FIXUP option is not specified in the OpenWRT build system for make check, a message is generated in the logs. In this GH Action, that message is filtered-in (see this line) as an error, and the build fails.

When the FIXUP option is specified, and in the Makefile we have PKG_MIRROR_HASH=skip, no log message regarding a hash check failure is generated, so with that this GH action is able to subsequently compile the package. This succeeds because the FIXUP option also fixes the Makefile to contain the correct value of PKG_MIRROR_HASH.

This PR adds the ability to specify the FIXUP option to make check (over here). I have also added a second, a more brute-force, option of IGNORE_PKG_HASH_ISSUE. This option simply disables the exit 1 statement when a package has a mismatched hash (over here).

I welcome suggestions for changes to this PR.

[qwerttvv]

let me add an example
https://github.com/qwerttvv/Router/blob/master/.github/workflows/smartdns.yml
https://github.com/pymumu/smartdns/blob/master/package/openwrt/Makefile

this actions compiles the latest smartdns

[cmangla]

I don't understand the motivation for this PR. Could you please explain and add it to the commit message?

@aparcar Done. I kept only the FIXUP option in this PR for simplicity and clarity.

[Ansuel]

@cmangla changes looks O.K. If you have extra time I would also check the diff in the package Makefile and if FIXUP enable also suggest the correct hash. Totally optional but would be welcome

An alternative solution might be "--nocheck" to disable package check entirely.

[EkkoG]

@Ansuel I mostly use FIXUP to fix the hash in Makefile, not use it to avoid build failure. I think this is the right way use it, so I thinks when FIXUP=1, the compile should not happen.

[cmangla]

@Ansuel I mostly use FIXUP to fix the hash in Makefile, not use it to avoid build failure. I think this is the right way use it, so I thinks when FIXUP=1, the compile should not happen.

@EkkoG Presumably, when you do use FIXUP to fix the hash in the Makefile, you do eventually compile it, right? In other words, the eventual goal of fixing the Makefile is to then launch a build with it. Given that, I do not see the point on insisting that the build should not happen when FIXUP=1.

Anyway, if the merge of this PR is predicated on this alternative mode of operation, please do let me know.

I do agree with @Ansuel though, that ideally the fixed Makefile should be presented so that the fixes can be retrieved and then the subsequent checks can pass without the FIXUP option.