524139 - eval .all.pp instead of including it

Status: RESOLVED FIXED

(Firefox Build System :: General, defect)

Description

[Benjamin Smedberg]

Currently we -include .all.pp with a FORCE rule to regenerate it. This sucks for build speed because we're restarting the parsing process. With gmake this means re-reading all the makefiles from disk. With pymake this means rereading at least .all.pp. In both cases this means redoing a lot of stats.

Comment 1

[Benjamin Smedberg]

Attachment: eval .all.pp instead of including it, rev. 1 [Checkin: Comment 7]

Ted, ISTR you had some framework for measuring build speed; although I have a strong hunch this will improve build speed it would be good to actually measure that before committing.

Comment 2

[(not currently active) Ted Mielczarek]

If by "framework" you mean "run a bunch of builds and time them", then yes. :)

Comment 3

[Ben Turner (not reading bugmail, use the needinfo flag!)]

Somehow this hangs my windows build early in configure right after it says:

creating js-config.h
js-config.h is unchanged
invoking make to create js-config script

Comment 4

[(not currently active) Ted Mielczarek]

I wonder how this ties in with bug 462463... (we should revisit that)

Comment 5

[(not currently active) Ted Mielczarek]

Comment on attachment 408039 [details] [diff] [review]
eval .all.pp instead of including it, rev. 1
[Checkin: Comment 7]

+$(patsubst %.cc,%.$(OBJ_SUFFIX),$(filter %.cc,$(CPPSRCS))): \

I think these static pattern rule changes belong in a different bug.

r=me with that part removed. I'll try to do a comparison build with and without this patch.

Comment 6

[[e-mail]]

(In reply to comment #3)
> Somehow this hangs my windows build early in configure right after it says:
> 
> creating js-config.h
> js-config.h is unchanged
> invoking make to create js-config script

My make doesn't like $(eval) inside an ifdef, so I had to move the
$(eval $(ALL_PP_RESULTS))
to just before the bug #### separator.

Comment 7

[(not currently active) Ted Mielczarek]

What version of make are you using?

Also, this landed:
http://hg.mozilla.org/mozilla-central/rev/29d7c945849d

Comment 8

[[e-mail]]

(In reply to comment #7)
> What version of make are you using?
It calls itself GNU Make 3.80

> Also, this landed:
> http://hg.mozilla.org/mozilla-central/rev/29d7c945849d
Exactly, but my gmake doesn't like the $(eval) inside an ifdef.

Comment 9

[Serge Gautherie (:sgautherie)]

Attachment: (Bv1-CC) Copy it to comm-central [Checkin: Comment 11+13]

Comment 10

[Justin Wood (:Callek)]

Comment on attachment 429504 [details] [diff] [review]
(Bv1-CC) Copy it to comm-central
[Checkin: Comment 11+13]

serge, this looks ok to land. Asking for feedback from Neil to see if there needs to be a solution for his gmake issue (or if he got it working right)

Comment 11

[Serge Gautherie (:sgautherie)]

Comment on attachment 429504 [details] [diff] [review]
(Bv1-CC) Copy it to comm-central
[Checkin: Comment 11+13]


http://hg.mozilla.org/comm-central/rev/7e6839c431c7

Comment 12

[[e-mail]]

Comment on attachment 429504 [details] [diff] [review]
(Bv1-CC) Copy it to comm-central
[Checkin: Comment 11+13]

Thanks for the heads-up; I'll have to apply the same workaround locally (move the eval to after all of the endifs).

Comment 13

[Serge Gautherie (:sgautherie)]

Comment on attachment 429504 [details] [diff] [review]
(Bv1-CC) Copy it to comm-central
[Checkin: Comment 11+13]


http://hg.mozilla.org/comm-central/rev/6c761f19ee0f
(Cv1-CC) Restore previous code on MOZILLA_1_9_2_BRANCH

NB: mddepend.pl part didn't land on m-1.9.2, so Thunderbird 3.1 busted :-<