make --with-extensions=all do what it says



Firefox Build System
17 years ago
4 months ago


(Reporter: dbaron, Assigned: dbaron)



Firefox Tracking Flags

(Not tracked)



(6 attachments)

I'd like to make --with-extensions=all more accurate, so that people
who are trying not to break extensions will actually build all working 
extensions when the try to do so.  This would mean that the tinderboxes
(on Ports) that build --with-extensions=all would be building these
extensions.  As a first pass, I'd like to add the extensions that 
clearly seem ready to be built -- those that are in --
part of --with-extensions=all.  That means adding inspector,
access-builtin, and transformiix to --with-extensions=all.

hewitt, aaronl, and peterv:  Is this OK with you?

I'll attach a patch once I add the REQUIRES lines...
It would be nice to fix the transformiix build system to use archives rather
than do everything (again) in extensions/transformiix/build/ first
Created attachment 30774 [details] [diff] [review]
patch to add access-builtin, inspector, and transformiix
Adding Axel Hecht, he knows much more about the Transformiix unix build system
than I do. David, could you expand a little on your last comment, I don't
understand what you mean. Also note that Transformiix is very close to being
turned on in the default build.

Comment 4

17 years ago
Adding inspector is fine with me.  It should build on win32, mac, and linux...
who knows what will happen on other platforms.

Comment 5

17 years ago
transformiix is blocked by the old makefiles. See 74879. 
Concerning David (recap of irc discussion): I don't like the idea
of static libs more or less than listing the object files twice.
"Due to historical reasons" bla applies.
We have a few dirs with just one .cpp, and my fingers curl up to my knees on the
idea of static libs for those. IMHO, it's not worth the effort converting.


Comment 6

17 years ago
r=cls on attach 30774.  I assume that changes for vixen, p3p, etc are in the works?

Comment 7

17 years ago
I'd would like the REQUIRES lines in transformiix to be ifdef MOZ_XSL. 
Currently, only the mozmodule depends on other stuff in mozilla, and we might
at one time have a dependency on string only for standalone.

aaronl said (on IRC) that this is fine with him.  So I just need to #ifdef those
REQUIRES lines...
Created attachment 31278 [details] [diff] [review]
revised patch for access-builtin, inspector, and transformiix

Comment 10

17 years ago

Comment 11

17 years ago
extensions=all would build transformiix standalone this way, but to build the
moz module, you need to set MOZ_XSL.
This should be synched/merged with 72141.

I tested that it works without --enable-xsl as well.  I'd like to see this
go in before bug 72141, so that we don't have to check it in urgently to
fix the bustage on senna.  (There are also some other issues in bug 72141
that seem to need to be addressed.)

Comment 13

17 years ago
David, if you've verified that it builds the transformiix module (not the
standalone), r=peterv. I'm surprised that that works.
No, it (without --enable-xsl) builds the standalone.  But I'd rather have this
in first and then fix bug 72141 (which would involve making it so that it really
builds the transformiix module).
I just checked in the bustage fixes (including the ifdef-ed REQUIRES lines for
transformiix) and I added access-builtin and inspector to --with-extensions=all,
but I didn't add transformiix yet, since it would currently build the standalone
(cls agreed that it should wait).  It should be added as part of bug 72141.
Created attachment 33311 [details] [diff] [review]
fix some weirdness in inspector makefiles that I just noticed

Comment 18

17 years ago
r=cls on the latest attach

Comment 19

17 years ago
LIB_SUFFIX is preferred over hardcoding .a
Fixed the error mkaply pointed out.  (For the record, the patch above fixed the
fact that we were creating "libinspector_s" and ".a" where we should have been
creating "libinspector_s.a", and the cause of the problem was a space at the end
of the LIBRARY_NAME line.)

I also had to make inspector link against gkgfx and libgkconshared_s to fix AIX
bustage.  The latter dependency really has to go.  The errors that led me to do
that were:

ld: 0711-317 ERROR: Undefined symbol: .NS_HexToRGB
ld: 0711-317 ERROR: Undefined symbol: .nsCSSProps::AddRefTable()
ld: 0711-317 ERROR: Undefined symbol: .nsCSSProps::ReleaseTable()
ld: 0711-317 ERROR: Undefined symbol: .nsCSSProps::GetStringValue(nsCSSProperty)
ld: 0711-317 ERROR: Undefined symbol: .nsCSSProps::LookupProperty(const nsAString&)

(I wonder why it was linking against rdfutil_s to begin with... that would be
good to remove as well.)

I also fixed some true->PR_TRUE and false->PR_FALSE bustage in inspector and one
file in access-builtin that had Windows line endings.
I filed bug 79091 about inspector's link-time dependencies.
Finishing this isn't at the top of my radar right now.  If someone else wants
to, go ahead.  The first bit was a good bit more work than I intended...
Priority: -- → P4
Target Milestone: --- → Future
I'll have a patch for the next step shortly.  This patch adds:
  ctl cview p3p universalchardet vixen

pics seems to require numerous files that aren't in the tree.
python seems complicated.

This patch adds new files for universalchardet and makes extensive
changes to p3p to get it building again (since it clearly hasn't been built for
quite a while).
Target Milestone: Future → mozilla0.9.4
Created attachment 45666 [details] [diff] [review]
second pass, version 1, diff -uN
Created attachment 45667 [details] [diff] [review]
second pass, version 1, diff -uwN
Created attachment 45676 [details] [diff] [review]
better p3p changes, incorporating ideas from bug 62399

Comment 28

17 years ago
r=cls on the makefile changes
harishd said by email that the p3p changes were fine with him

yokoyama said adding the universalchardet makefiles was fine (and noted bug
92341, which says it should be on by default)

I still haven't heard back from katakai (for ctl), rginda (for cview), or ben
(for vixen).
I'm actually thinking I probably shouldn't add p3p since it adds significant
leaks to startup and that will make the leak stats more confusing.

Comment 31

17 years ago
There's a separate bug for ctl, bug 95258.

Comment 32

17 years ago
Target Milestone: mozilla0.9.4 → mozilla0.9.5

Comment 33

17 years ago
*** Bug 91305 has been marked as a duplicate of this bug. ***
Target Milestone: mozilla0.9.5 → mozilla0.9.6
Target Milestone: mozilla0.9.6 → Future
Priority: P4 → P5

Comment 34

15 years ago
*** Bug 221233 has been marked as a duplicate of this bug. ***

Comment 35

15 years ago
I'm not sure that Bug 221233 is actually a duplicate of this one, since cview is
actually included in --with-extensions=all in configure, but it just doesn't get
compiled. (Or is that what this bug is about?)
Product: Browser → Seamonkey
Are you still working on this ?
Component: Build Config → Build Config
Product: Mozilla Application Suite → Core
And as bug 450015 notes, neither was anyone else. At all. Ever.
Last Resolved: 10 years ago
Resolution: --- → WONTFIX


4 months ago
Product: Core → Firefox Build System
You need to log in before you can comment on or make changes to this bug.