227896 - Composer build issues with Win32 & MinGW GCC

Status: VERIFIED FIXED

(SeaMonkey :: Build Config, defect)

Description

[David G King]

I have seen a couple of issues with building Composer on WIN32 with GCC, I will
be describing and attaching patches to this bug.

Comment 1

[David G King]

First problem is that NS_ALERTSERVICE_CONTRACTID is undefined in
mozilla/xpfe/components/build2/nsModule.cpp

Comment 2

[David G King]

Attachment: Patch for first problem

Had to patch Makefile.in and add an include into nsModule.cpp

Comment 3

[David G King]

Second problem is about the infamous Win32 gcc call to ::main problem.

Comment 4

[David G King]

Attachment: Patch for second problem

The "normal" removal of the ::Main problem. Also, found that this bug is
dependant on an earlier fix (from cls) in bug #220978

Comment 5

[David G King]

Attachment: Second attempt at first problem

Well, the first patch fixed Composer/Sunbird, but Thunderbird wasn't happy with
it. This new patch works with both.

Comment 6

[David G King]

Comment on attachment 137173 [details] [diff] [review]
Second attempt at first problem

OK, lets see if this patch will pass a review?

Comment 7

[Brian Ryner (not reading)]

Hm, no, this isn't right.  We should switch the new-toolkit products to use the
alerts service in toolkit/components.  I see thunderbird is still using the one
from xpfe.  Scott, any thoughts?

Comment 8

[David G King]

For more information, which may not be relevant, may I suggest a quick look at
Bug #194315, and especially 

http://bugzilla.mozilla.org/show_bug.cgi?id=194315#c10

where it was decided to create the "build2" dir under XPFE for Minotaur sync
with the Trunk.

Comment 9

[Brian Ryner (not reading)]

Comment on attachment 137173 [details] [diff] [review]
Second attempt at first problem

review request -> mscott, since he can better address  the questions here. 
Scott, please read comment 7 before reviewing this.

Comment 10

[David G King]

Comment on attachment 137099 [details] [diff] [review]
Patch for second problem

Asking cls for a review of this patch as I copied it from one of his patches
and he knows GNU on WIN32 better than I do.

Comment 11

[David G King]

Comment on attachment 137099 [details] [diff] [review]
Patch for second problem

Asking Brendan for an SR...keeping in mind that I have no checkin rights.

Comment 12

[Brendan Eich [:brendan]]

Comment on attachment 137099 [details] [diff] [review]
Patch for second problem

You don't generally need sr for build changes, so if the module owner is ok,
and if you have no doubts about conventions to use (e.g., __GNUC__ not
defined), go for it.

Almost a matter of taste issue: maybe combine the #if{def,ndef} lines and
#endifs via #if defined XP_WIN && !defined __GNUC__ ?  A comment about MinGW
defining both would help the newcomers.

/be

Comment 13

[David G King]

Hmmm, I agree with the reformating of the IFDEF's. And I have added a comment
about why this is bad for MingW. New patch coming up once I've tested it...well,
at least made sure it compiles.

As for the SR, I always ask for them at the moment, as my Mozilla knowledge is
rather low. Once I gain more confidence, that will change.

Comment 14

[David G King]

Attachment: Final patch for second problem

New patch as per comments from SR

Comment 15

[David G King]

Comment on attachment 139704 [details] [diff] [review]
Final patch for second problem

Carrying over R and SR from prior patch as only minor chnages made.

Comment 16

[cls]

Comment on attachment 139704 [details] [diff] [review]
Final patch for second problem

Patch for the second problem has been checked in.

Comment 17

[David G King]

Thanks cls, marking Fixed.

Comment 18

[Brian Ryner (not reading)]

Comment on attachment 137173 [details] [diff] [review]
Second attempt at first problem

clearing old request

Comment 19

[David G King]

V'ing