569782 - Packaging changes for Gecko 1.9.3

Status: RESOLVED WONTFIX

(Camino Graveyard :: General, defect)

Description

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

There was a bunch of jar collapsing going on on 1.9.3; we'll need to remove some res packaging, remove comm.jar (if I don't wack it first), and probably move our missing/loading images into being packaged by pinstripe (I believe they're res/ files).

Comment 1

[Stuart Morgan]

Also JavaEmbeddingPlugin.bundle and MRJPlugin.plugin.

Comment 2

[Stuart Morgan]

Attachment: Decaf

This takes care of the two Java plugins, since that's easy :)

Comment 3

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Attachment: Fix

This takes care of everything mentioned in philippe's log, plus moving some things around in pinstripe to adapt to Gecko changes.

One thing we need to watch out for: hiddenWindow.html is now packaged unconditionally in the jar, which means suddenly it'll exist (the chrome URL will resolve) for Camino.  IIRC, a bunch of widget code uses the presence/absence of hiddenWindow to check for embedding vs. XUL; depending on how those checks are written (or how fuzzy my memory is), we might have to strip hiddenWindow.html out of the jar before packaging.

I haven't tried to build with this yet, but I believe we still fail before we get to this point, anyway.

Comment 4

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

(In reply to comment #3)
> This takes care of everything mentioned in philippe's log: http://pastebin.mozilla.org/730918

Comment 5

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Last comment here for now, I promise: there may be libs and xpts that we have to *add* to get new Gecko features, or js(m)s to make current stuff work, etc.; this should just solve the immediate build failure problems.

Comment 6

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Debug builds need a few more fixes:

	PBXCp build/Debug/Camino.app/Contents/MacOS/components/libucvmath.dylib ../dist/bin/components/libucvmath.dylib
	PBXCp build/Debug/Camino.app/Contents/MacOS/components/libcaps.dylib ../dist/bin/components/libcaps.dylib
	PBXCp build/Debug/Camino.app/Contents/MacOS/components/libxpconnect.dylib ../dist/bin/components/libxpconnect.dylib
	PBXCp build/Debug/Camino.app/Contents/MacOS/components/liboji.dylib ../modules/oji/src/liboji.dylib
	PBXCp build/Debug/Camino.app/Contents/MacOS/components/libcaminosafebrowsing.dylib ../dist/bin/components/libcaminosafebrowsing.dylib
	PBXCp build/Debug/Camino.app/Contents/MacOS/libjsj.dylib ../dist/bin/libjsj.dylib

I'll have to track down what happened to them (oji is certainly gone, but some of the others seem like core Gecko functionality), especially why libcaminosafebrowsing.dylib failed to build.

Comment 7

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

(In reply to comment #6)
> especially why
> libcaminosafebrowsing.dylib failed to build.

That's me not having learned my lesson about commenting out things in Makefile continuation lines :P

Comment 8

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

(In reply to comment #6)
> Debug builds need a few more fixes:
> 
>     PBXCp build/Debug/Camino.app/Contents/MacOS/components/libucvmath.dylib
> ../dist/bin/components/libucvmath.dylib

Gone, bug 483116.

>     PBXCp build/Debug/Camino.app/Contents/MacOS/components/libcaps.dylib
> ../dist/bin/components/libcaps.dylib
>     PBXCp build/Debug/Camino.app/Contents/MacOS/components/libxpconnect.dylib
> ../dist/bin/components/libxpconnect.dylib

Linked into layout, bug 560199.

Both of those will need nsStaticComponents.cpp changes.

>     PBXCp build/Debug/Camino.app/Contents/MacOS/libjsj.dylib
> ../dist/bin/libjsj.dylib

That's liveconnect, so also gone.  Either it or OJI is probably also in nsStaticComps.

New patch in a bit.

Comment 9

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Attachment: Fix, v1.1

This removes oji, jsj, ucvmath {a|dylib}, since those are totally gone from the tree.  

It also removes caps and xpconnect dylibs, since they're now linked into layout; I'd need to check a static log carefully to see if caps.a and xpconnect.a generate ld errors once the Camino binary itself builds. I think they're probably gone, too, but erring on the side of caution here for now.

Also, nsStaticComponents removals for caps (nsSecurityManagerModule), xpconnect, ucvmath, and oji (nsCJVMManagerModule).

Comment 10

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Attachment: Fix, v1.2

Now with more mozalloc and ycbcr!  This succeeds for debug/shared (and crashes in embedding code); might still have to wack it for static (will have to fix ycbcr for static; it'll be a static lib of some sort).

Comment 11

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Attachment: crashlog

(In reply to comment #10)
> (and crashes in embedding code)

First signs of warning are the 

nsNativeModuleLoader::LoadModule("/Users/smokey/Camino/dev/m-c/CaminoTrunk/camino/build/Debug/Camino.app/Contents/MacOS/components/libgklayout.dylib") - load FAILED, rv: 80004005, error:
	Unknown error: -2804

failing twice(!) during startup (compreg is happening twice, sequentially!)

followed by that murph-early-warning sign, 

Error initializing SafeBrowsingListManager: Failed to get the '@mozilla.org/url-classifier/listmanager' service

I assume layout failing makes everything unhappy, but maybe there's something even lower-level than that that we can't see.

Comment 12

[Stuart Morgan]

Summarizing from irc, we need to package:
libgfxutils.dylib
Services.jsm
PlacesUtils.jsm

At that point we hit the bug 561957 crash in widget code, which suggests we may be at the end of the packaging rabbit hole (for shared; I haven't tried static).

Comment 13

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Attachment: Fix, v1.3

This has the new shared fixes for gfxutils, and will fix the static build, too (ycbcr and gfxutils added, OS frameworks linked [for gfxutils], caps and xpconnect removed), at least as far as linking.  It also adds the .jsms in both places.  

We'll still need to audit xpts for new things added as parts of existing modules/libs.

One note: gfxutils notes specifically that it's not a component; we don't really have a place for "static things that aren't static all over, and which aren't components", so it's in Static Components.  This bothers me a bit, but I'm not sure I want to unravel the mess that is Gecko right now.

Comment 14

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

> At that point we hit the bug 561957 crash in widget code, which suggests we may
> be at the end of the packaging rabbit hole (for shared; I haven't tried
> static).

I tried to launch the static build, and it basically killed my Mac while crashing from bug 561957.  DO NOT TRY THIS AT HOME!

Comment 15

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Attachment: Fix, v1.3a

This just updates the patch to account for the bitrot caused by bug 570361.

Comment 16

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Once bug 569195 lands, we'll need to package mathfontSTIXSizeOneSym.properties, too.

(In reply to comment #15)
> Created an attachment (id=449542) [details]
> Fix, v1.3a
> 
> This just updates the patch to account for the bitrot caused by bug 570361.

There's more bitrot from the second patch for bug 570361; I'd like to get bug 570373 landed before I spin another patch, to keep the patch-churn here in check (and the merging here is trivial, if you know what's being changed).

Comment 17

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Attachment: Fix, 1.3b

Here's the rest of the bitrot from bug 570361 merged.

Comment 18

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Attachment: Fix, 1.3c

This adds the new mathfont properties file from bug 569195 (and actually packages libgfxutils; before we just linked against it without packaging :( ).

Comment 19

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

"Inability to build against Gecko 2 (FKA 1.9.3) is permanent."