780357 - Port bug 776968, bug 774032, bug 780159, bug 776561, bug 778740 to comm-central

Status: RESOLVED FIXED

(MailNews Core :: Build Config, defect)

Description

[Philipp Kewisch [:Fallen][📱📆]]

I just saw a pretty large push go to inbound that changes a lot of things in synchronized files like configure.in. Someone should go through this stuff to see what is needed for comm-central.

bug 776968 - Avoid modifying autoconf.mk after running nspr configure
bug 774032 (8 parts!)
  - Avoid Preprocessor.py replacing undefined variables with the 
    attemptSubstitution filter
  - Allow to disable markers in Preprocessor.py
  - Remove variables with no AC_SUBST in autoconf.mk.in
  - Add check-sync-dirs exception for *.pyc under build/
  - Use explicit _NON_GLOBAL_ACDEFINES, and rely less on pattern matching
  - Replace autoconf handling of config files and headers with our own
  - Remove make-makefile and acoutput-fast.pl
  - bonus: Use @DEPTH@ and @relativesrcdir@ in Makefile.in
bug 780159 - Remove obsolete REGCHROME
bug 776561 - Avoid OS_LIBS containing -lrt when there is no config.cache
bug 778740 - Allow to use Linux kernel headers for perf events

Comment 1

[Philipp Kewisch [:Fallen][📱📆]]

glandium says we should only need bug 774032 and bug 780159. About bug 776968 he says it should change nothing since there was a workaround to avoid reconfigures.

Comment 2

[Mike Hommey [:glandium]]

bug 778740 is irrelevant for c-c because it doesn't use linux perf event stuff.
bug 776561 is irrelevant for c-c because it doesn't have the check for clock_gettime that added -lrt on m-c
bug 776968 may actually be good to port, for the part that *is* in c-c (the autoconf.mk rewriting part is not), because it avoids using shell commands each time NSPR_LIBS is used.

Comment 3

[Mike Hommey [:glandium]]

Attachment: Partial port of bug 774032

This should port the following parts of bug 774032:
  - Remove variables with no AC_SUBST in autoconf.mk.in
  - Use explicit _NON_GLOBAL_ACDEFINES, and rely less on pattern matching
  - Replace autoconf handling of config files and headers with our own
but it doesn't contain the following files that need to be copied from m-c:
  - build/ConfigStatus.py
  - config/Expression.py
  - config/Preprocessor.py

The following patches are irrelevant if you just copy the last versions of the above files:
  - Avoid Preprocessor.py replacing undefined variables with the 
    attemptSubstitution filter
  - Allow to disable markers in Preprocessor.py

This is not required, because it's already in c-c, somehow:
  - Add check-sync-dirs exception for *.pyc under build/

This is not required, because the files are not in c-c:
  - Remove make-makefile and acoutput-fast.pl

This is not strictly required but is a nice bonus (but not in this patch):
  - bonus: Use @DEPTH@ and @relativesrcdir@ in Makefile.in

Comment 4

[Mike Hommey [:glandium]]

Also, the part for WIN_TOP_SRC is not ported. It may require porting things from c-c to m-c (cygwin parts)

Comment 5

[Mike Hommey [:glandium]]

Attachment: Partial port of bug 774032

Refreshed with bug 780446 and bug 780421

Comment 6

[Mike Hommey [:glandium]]

Note that you'll also want bug 763987 (the goal being to prepare autoconf.mk for bug 742795, and for that, it needs all variables to be in the form VAR = @VAR@, not VAR = FOO @VAR@ or VAR = @VAR2@)

Comment 7

[Mike Hommey [:glandium]]

(In reply to Mike Hommey [:glandium] from comment #6)
> Note that you'll also want bug 763987 (the goal being to prepare autoconf.mk
> for bug 742795, and for that, it needs all variables to be in the form VAR =
> @VAR@, not VAR = FOO @VAR@ or VAR = @VAR2@)

That's however not needed to unbreak the build on c-c.

(In reply to Mike Hommey [:glandium] from comment #4)
> Also, the part for WIN_TOP_SRC is not ported. It may require porting things
> from c-c to m-c (cygwin parts)

cygwin support was removed from m-c in bug 462361. I guess we can just remove the cygwin part in configure.in in c-c.

Comment 8

[Mike Hommey [:glandium]]

Attachment: Allow m-c's ConfigStatus.py to be used from c-c

This allows c-c to configure properly with the PoC I already attached, and without copying files from m-c.

Comment 9

[Mike Hommey [:glandium]]

Comment on attachment 649091 [details] [diff] [review]
Partial port of bug 774032

I think this is enough, with the m-c patch i attached, to unbreak c-c.

Comment 10

[Mike Hommey [:glandium]]

Landed the m-c part:
https://hg.mozilla.org/integration/mozilla-inbound/rev/91de8ff323d4

Comment 11

[Mike Hommey [:glandium]]

And the missing sync with js/src:
https://hg.mozilla.org/integration/mozilla-inbound/rev/949770676b0f

Comment 13

[Robert Kaiser]

As this is about the general comm-central build system and not Thunderbird-specific, it belongs in MailNews Core

Comment 14

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/91de8ff323d4
https://hg.mozilla.org/mozilla-central/rev/949770676b0f

Comment 15

[Andreas Nilsson (:andreasn)]

I've experienced this bug on both Linux and Windows(7) so changing platform from OSX to All.

Comment 16

[Mike Hommey [:glandium]]

Landed on c-c to unbreak it.
http://hg.mozilla.org/comm-central/rev/493c0d246a5e

Maybe the remaining parts should be tracked in a separate bug, and we should close this one?

Comment 17

[Mark Banner (:standard8)]

Thanks for your work on this Mike.

I agree with Mike, we should do follow-ups in a separate bug. Can someone formulate a list please?

Comment 18

[Mike Hommey [:glandium]]

(In reply to Mark Banner (:standard8) from comment #17)
> Thanks for your work on this Mike.
> 
> I agree with Mike, we should do follow-ups in a separate bug. Can someone
> formulate a list please?

Filed bug 781446.