212509 - nsToolkitCompsCID.h: No such file or directory (and numerous firebird build failures)

Status: VERIFIED FIXED

(Firefox Build System :: General, defect)

Description

[John Levon]

CVS trunk as start of today 2003-07-12. Compile errors out with :

nsExternalHelperAppService.cpp:91:31: nsToolkitCompsCID.h: No such file or directory

in uriloader/exthandler. Linux RH 8.0. Used "make -f client.mk checkout clobber
build_all". .mozconfig :

export MOZ_PHOENIX=1
mk_add_options MOZ_PHOENIX=1
export MOZ_THUNDERBIRD=1
mk_add_options MOZ_THUNDERBIRD=1
ac_add_options --disable-installer
ac_add_options --disable-activex
ac_add_options --disable-activex-scripting
ac_add_options --disable-profilesharing
ac_add_options --enable-extensions=wallet
ac_add_options --enable-necko-protocols=http,file,jar,viewsource,res,data
ac_add_options --enable-image-decoders=png,gif,jpeg
ac_add_options --disable-accessibility
ac_add_options --with-extensions="default"
ac_add_options --enable-xft
ac_add_options --disable-bidi
ac_add_options --disable-ldap
ac_add_options --disable-postscript
ac_add_options --disable-dtd-debug
ac_add_options --disable-debug
ac_add_options --disable-tests
ac_add_options --disable-mailnews
ac_add_options --disable-composer
ac_add_options --disable-logging
ac_add_options --enable-crypto
ac_add_options --enable-optimize="-O2 -g -pipe -march=i686"
ac_add_options --disable-jsd
ac_add_options --disable-venkman
mk_add_options --mcpu=i686
mk_add_options --march=i686

Comment 1

[Matthias Versen [:Matti]]

-> FB

Comment 2

[John Levon]

Attachment: build fix

The file doesn't even use nsToolkitCompsID.h - remove the include altogether

Comment 3

[Alex Stewart]

*** Bug 212909 has been marked as a duplicate of this bug. ***

Comment 4

[Alex Stewart]

So after a bit of leg...err...finger-work, I've come up with a hypothesis for a
(really short) timeline.

1) Christian Biesinger fixes bug 124014 and makes links handled by an external
application show up in history.  He adds the line
nsCOMPtr<nsIGlobalHistory> history(do_GetService(NS_GLOBALHISTORY_CONTRACTID));
which is the first time that NS_GLOBALHISTORY_CONTRACTID is used in
nsExternalHelperAppService.cpp.  He adds the #include "nsIGlobalHistory.h" to
bring in the symbol.  This is revision 1.192 of NSExternalHelperAppService.cpp

2) David Baron checks in revision 1.193, which adds the #include of
nsToolkitCompsCID.h if Firebird is being built.  According to the build comment,
 this fixes MOZ_PHOENIX bustage.

This all happened on June 30th.  There probably is a step missing, since
nsIGlobalHistory.idl lives in the xpfe directory.


So now the pill is worse than the original illness.  With levon's patch, I can
build the uriloader.  Since I'm not qualified to give out "r="s, I'll give it a
t=am2stewa for "tested and seems to work".

(Oh, and Hardware -> All, since it happens on Sun/Solaris as well.  (There
really needs to be a OS: Unices option if OS is going to be relevant.))

Comment 5

[John Levon]

Attachment: newer patch

Some more build fixes !

Comment 6

[John Levon]

Just attached a newer patch with some other places where the build fails.

It seems that nobody is interested in building firebird at the moment. Nobody 
on IRC, or the forums, or bugzilla, seems to care.

Unfortunately I'm now stuck on :

nsAppRunner.cpp:86:35: InstallCleanupDefines.h: No such file or directory
nsAppRunner.cpp:87:31: nsISoftwareUpdate.h: No such file or directory
...
gmake[4]: Leaving directory `/home/moz/src/mozilla/toolkit/xre'

xpinstall (where these files are from) is listed in xre's Makefile.in REQUIRES list

Is there anybody out there ?

Comment 7

[John Levon]

Doing a "make" in xpinstall fixed the last problem. I guess I could
turn on xpinstall dir if MOZ_PHOENIX in the toplevel Makefile.in but that just
feels like it's wrong to me...

Continuing with the build ...

Comment 8

[Alex Stewart]

After more investigation (reasoning that there must be one root cause behind all
these build failures), I've discovered bug 107302, which deals with breaking
stuff into tiers.

On July 15th, cls checked in a patch to create a tier_50, which currently
contains tools/jprof, xpfe, themes, tools/leaky, tools/codesights, l10n, and
toolkit.  This means that the required headers from toolkit aren't installed
when tier_9 (where all the good stuff lives) is built.  So we may have to add
some directories from toolkit to tier_9, kinda like lines 217-221 of
mozilla/Makefile.in (which add some directories from xpfe).  That way, their
header files will be in dist/include when they're needed.

Comment 9

[John Levon]

fyi, with the given patch, and the manual make in xpinstall, it finally
built and worked fine. Cc'd cls in the hope somebody can resolve the tier
problem properly and apply the rest of the patch.

Comment 10

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Firbird forked a bunch of interfaces, some of which had CIDs in the IDL files,
and moved the CIDs into nsToolkitCompsCID.h.  We shouldn't be using two separate
IDL files for the same interface in the build at different points, so both
patches here look wrong.  I suspect seawood's changes caused the xpfe versions
of the interfaces to be used during part of the build process.

It would be easiest to un-fork these files.  Have there been changes other than
moving the CIDs out of the IDL?

Comment 11

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

(The approach in comment 8 would be a sufficient workaround.  When that's done,
though, we also should make the additions to tier_9 that weren't in Firebird
before |ifndef MOZ_PHOENIX|.

Comment 12

[Alex Stewart]

Attachment: Changes to Makefile.in to get Firebird to compile

Doesn't touch any file but Makefile.in.  Since I made my changes by
trial-and-error, I'd definitely want other people to try this patch and report
back.  Oh, and it doesn't require the previous patch to be applied.

Comment 13

[John Levon]

> The file doesn't even use nsToolkitCompsID.h - remove the include altogether

I don't understand how this part of the patch can be wrong ? Why include
an unused header ?

Comment 14

[Scott MacGregor]

Thunderbird builds are also broken with bustage in mozilla\toolkit. Is this
bustage related to what you are seeing in firebird?

illa-config.h -Wp,-MD,.deps/nsAppRunner.pp
/home/andre/devel/cvs/mozilla/toolkit/xre/nsAppRunner.cpp
/home/andre/devel/cvs/mozilla/toolkit/xre/nsAppRunner.cpp:86:35:
InstallCleanupDefines.h: No such file or directory
/home/andre/devel/cvs/mozilla/toolkit/xre/nsAppRunner.cpp:87:31:
nsISoftwareUpdate.h: No such file or directory
/home/andre/devel/cvs/mozilla/toolkit/xre/nsAppRunner.cpp: In function `
   nsresult VerifyInstallation(int, char**)':
/home/andre/devel/cvs/mozilla/toolkit/xre/nsAppRunner.cpp:980: error: `
   CLEANUP_REGISTRY' undeclared (first use this function)
/home/andre/devel/cvs/mozilla/toolkit/xre/nsAppRunner.cpp:980: error: (Each
   undeclared identifier is reported only once for each function it appears
   in.)
/home/andre/devel/cvs/mozilla/toolkit/xre/nsAppRunner.cpp:989: error: `
   CLEANUP_MESSAGE_FILENAME' undeclared (first use this function)
/home/andre/devel/cvs/mozilla/toolkit/xre/nsAppRunner.cpp:993: error: `
   CLEANUP_UTIL' undeclared (first use this function) 

Comment 15

[Roland Mainz]

Mhhh... does it make sense to add the keyword "smoketest" (blocker) ? =:-)

Comment 16

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

> > The file doesn't even use nsToolkitCompsID.h - remove the include altogether
> 
> I don't understand how this part of the patch can be wrong ? Why include
> an unused header ?

Because you're wrong that it's unused.  It uses |NS_GLOBALHISTORY_CONTRACTID|.

Comment 17

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Comment on attachment 127989 [details] [diff] [review]
Changes to Makefile.in to get Firebird to compile

>+# Common dirs

No need for a comment saying something that's obvious from looking at the code.

> tier_9_dirs	+= \
> 		xpfe/appshell \
>-		xpfe/browser/public \
>-		xpfe/components/history/public \
> 		xpfe/components/sidebar/public \
> 		xpfe/components/xremote/public \
>+		xpfe/browser/public \
>+		$(NULL)

No need to move xpfe/browser/public -- it just makes the diff bigger.

>+ifdef MOZ_XUL_APP
>+tier_9_dirs	+= \
>+		toolkit/components/autocomplete \
>+		toolkit/components/history \
>+		toolkit/components/satchel \
>+		toolkit/components/build \
>+		browser/components/bookmarks \
>+		$(NULL)

Why do you need autocomplete, satchel, and bookmarks?  I'd think you'd only
need history and build.

>+else
>+tier_9_dirs	+= \
>+		xpfe/components/history/public \
> 		$(NULL)
>+endif

Comment 18

[Alex Stewart]

Scott: Sure looks like it.  Version 1.222 of Makefile.in moved xpinstall from
tier_9 to tier_90.  I don't know why I didn't have to deal with that when I made
my patch.  It sure looks like toolkit uses the exports from xpinstall, thus it
needs to be in the same tier.  (i.e. my patch is incomplete)

Anyone else for reverting version 1.222 until the different parts of Mozilla
aren't tied so closely to each other?

As for the smoketest blockingness of this bug, it blocks clobber .*bird builds.
 Is that good enough for smoketest?

Comment 19

[Scott MacGregor]

David said:
"Why do you need autocomplete, satchel, and bookmarks?  I'd think you'd only
need history and build."

Please don't add that to MOZ_XUL_APP. Those should be MOZ_PHOENIX ifdef
makefiles. MOZ_XUL_APP is for the suite of stand alone applications including
firebird and thunderbird. Thunderbid does not use satchel, the new auto complete
or bookmarks. Moving it into the ifdef makes it depend on those modules. That's
what MOZ_PHOENIX is for. =)

Good catch dbaron.

Comment 20

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Oh, I see the problem.  The build directory doesn't have a separate public
directory, so this tries to put it in tier 9.  That's a problem.

We need to un-fork nsIGlobalHistory anyway, since it's frozen.

I'll propose a patch shortly.

Comment 21

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: patch

This unforks the *contents* of nsIGlobalHistory and fixes the root Makefile.in.
 I didn't completely clobber, but I did clobber all the files that included
nsToolkitCompsCID.h.

Comment 22

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

I also needed to make this change to mozilla/Makefile.in:

 ifdef MOZ_XUL_APP
 tier_50_dirs   += toolkit
+endif
+
+ifdef MOZ_PHOENIX
+# xpfe/components/search depends on nsIBookmarksService.idl
+tier_50_dirs    += browser/components/bookmarks/public
 endif

and also needed to add some additional missing makefiles to allmakefiles.sh.

Comment 23

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: patch

I already checked in the allmakefiles.sh changes, since they're obvious. 
Here's the rest of the patch for review.

Comment 24

[Scott MacGregor]

Comment on attachment 128082 [details] [diff] [review]
patch

r=mscott if you need it.

Comment 25

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Fix checked in to trunk, 2003-07-19 14:19 -0700.

Comment 26

[Alfons Hoogervorst]

building Firebird under Linux still breaks on my system, now in tier_90,  
with an invalid shell command line - out the top of my head  
 
	/usr/bin/sh for d in ; XXXXX done;  
	(with XXXX being some command) 
 
Empty for...loop. 
 
The last commit I had was this one 
http://bonsai.mozilla.org/cvsview2.cgi?subdir=mozilla/security/coreconf&files=Linux2.6.mk&command=DIRECTORY&branch=HEAD&root=/cvsroot 
 

Comment 27

[Alfons Hoogervorst]

Here's the error message itself: 
 
--8<-- 
make[3]: Leaving directory `/alfons/Projects/mozilla/xpinstall' 
make[2]: Leaving directory `/alfons/Projects/mozilla' 
/usr/bin/make tier_90 
make[2]: Entering directory `/alfons/Projects/mozilla' 
tier_90: 
/bin/sh: -c: line 1: syntax error near unexpected token `;' 
/bin/sh: -c: line 1: `set -e;   for d in ; do   /usr/bin/make -C $d export;  
done ;  /usr/bin/perl ./config/purge-old-headers.pl ./dist/include ; for d in 
; do   /usr/bin/make -C $d libs;  done' 
make[2]: *** [tier_90] Error 2 
make[2]: Leaving directory `/alfons/Projects/mozilla' 
make[1]: *** [default] Error 2 
make[1]: Leaving directory `/alfons/Projects/mozilla' 
make: *** [build] Error 2 
-->8-- 

Comment 28

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: possible patch

Does this fix that error.

This was the change I had to make that I was least happy about.  However, I
think the intent of toolkit/xre/ was really that it be something that
applications should build on, so I didn't want to split that into tier_99, but
maybe that's the better solution?

Comment 29

[Matthias Versen [:Matti]]

Comment on attachment 128024 [details] [diff] [review]
patch

clearing obsolete review request

Comment 30

[Alfons Hoogervorst]

yes, that fixes it. thanks.

Comment 31

[cls]

Comment on attachment 128088 [details] [diff] [review]
possible patch

That's fugly.  Just move xpinstall back into tier_50 unconditionally.  We don't
need different build paths for basic components based upon the app being built.

Comment 32

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: patch

Comment 33

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Checked in.

Comment 34

[Simon Paquet [:sipaq]]

Verified