857812 - Reimplement configure command

Status: RESOLVED DUPLICATE of bug 865445

(Firefox Build System :: General, defect)

Description

[Steve Fink [:sfink] [:s:]]

I want it because I want to be able to do a fresh checkout and build a portion
of the tree. Currently, this would fail because the objdir has not yet been
created, and can only be created with a full ./mach build.

For my specific case, this could also be implemented by making

  ./mach build js

automatically do a configure first instead of erroring out. But I think a
general configure command is more useful.

Comment 1

[Gregory Szorc [:gps]]

For the first revision we can just proxy this through |make -f client.mk configure| like we do for |mach build|. Should be pretty easy.

Comment 2

[Steve Fink [:sfink] [:s:]]

Attachment: Implement |mach configure|

Yeah, I was thinking it would look something like this, but maybe prettier.

Comment 3

[Gregory Szorc [:gps]]

Comment on attachment 733090 [details] [diff] [review]
Implement |mach configure|

Review of attachment 733090 [details] [diff] [review]:
-----------------------------------------------------------------

Oh, this is an interesting approach. I assumed a |mach configure| would be a completely new command that simply invokes |make -f client.mk configure|. I do like parts of this patch (such as automatically running configure if the tree isn't built). However, I'm trying to decide if there are any benefits to reusing the |mach build| logic to launch configure over just a a real simple command that always launches |make -f client.mk configure|. I'm not convinced there are. Why not just have a new function that simply calls self._run_make against client.mk?

Comment 4

[Steve Fink [:sfink] [:s:]]

I mostly did it that way to avoid figuring out what parts of the build function I would need to share with the configure command. Mainly, that means all of the warnings processing stuff -- do I want that or not? I assume the slow finder stuff isn't applicable.

Also, what about

        # This code is only meant to be temporary until the more robust tree
        # building code in bug 780329 lands.

Given that bug 780329 has landed, is this just obsolete or did that bug end up not changing whatever it was?

Comment 5

[Gregory Szorc [:gps]]

Comment on attachment 733090 [details] [diff] [review]
Implement |mach configure|

Review of attachment 733090 [details] [diff] [review]:
-----------------------------------------------------------------

You shouldn't need significant overlap between |mach build| and |mach configure|. Let's just have |mach configure| self._run_make() with client.mk's configure target.

Also, yes, that comment is stale.