83393 - gtkEmbed: re-add psm to gtkEmbed build package

Status: VERIFIED FIXED

(Core Graveyard :: Embedding: APIs, defect)

Description

[Chris McAfee]

from bug 82141, we had to yank psm from gtkEmbed to get
around a crasher problem.  We need to go put psm back
and make this work.

Comment 1

[Adam Lock]

I have detected several problems with the embedding manifest which prevent
TestGTKEmbed (much saner than gtkEmbed) from working.

1. The help module seems to be a default extension nowadays. For some reason,
opening a page with a certificate seems to want to open help.js/xul/css etc so
you need them in the manifest if help is built.
2. Classic skin changes. A number of classic skin files such as button.css
appear to be missing from the embed manifest. PSM can't pose a XUL dialog
properly without the skin.

Comment 2

[Adam Lock]

Attachment: Patch more or less makes ssl functional in embedding (notes follow)

Comment 3

[Adam Lock]

The patch enables PSM for embedding by uncommenting the psm components and
updating the embedding jar manifest. Basically the update consists of using the
current of list of what goes in skin/classic/global, content/global and a little
bit of skin/classic/communicator.

All my testing was done with TestGtkEmbed since I know that is reasonably sane.
With the new embed.jar, TestGTKEmbed poses warning dialogs and the certificate
info and then loads the https (https://www.fortify.net/sslcheck.html)

There is still one problem remaining in the chrome which is some language
strings for OK, Cancel are not found so some buttons are blank. This is probably
some missing files.

I will have a go tomorrow at gettting gtkEmbed to work and fixing the missing
strings.

Comment 4

[Adam Lock]

Attachment: Updated patch sorts out string problems.

Comment 5

[Chris McAfee]

tried this patch, TestGtkEmbed seems fine, gtkEmbed doesn't
do anything with the fortify URL but doesn't crash.

Comment 6

[Chris McAfee]

*** Bug 81884 has been marked as a duplicate of this bug. ***

Comment 7

[Chris McAfee]

ah, thanks to bryner for a clue: NSS won't initialize without
a profile directory. I comment out the profile initialization
line in TestGtkEmbed.cpp and see the same blank white page that
gtkEmbed shows with the fortify url.

Comment 8

[Chris McAfee]

Attachment: Patch that initializes a profile for gtkEmbed

Comment 9

[Chris McAfee]

Attached patch borrowed from TestGtkEmbed app, this initializes
a profile in ~/.gtkEmbed/gtkEmbed.  The fortify URL now
brings up a cert "dialog" (another gtkEmbed-sized window) and
then things digress into JS window errors.  It looks like psm
is loaded and at least trying to do something at least.

Comment 10

[Adam Lock]

Attachment: Perl script to automatically generate manifest

Comment 11

[Adam Lock]

The attached perl script should hopefully make it easier to keep the embedding 
chrome manifest up to date rather than checking in new versions of 
embed-jar.mn each time. The script reads the list of directories and files 
from %embed_files and builds the manifest from that. We could probably 
incorporate it into the makefile, but for the moment it must be run manually.

Usage:

./gen_mn.pl -mozpath /usr/src/mozilla > embed-jar.mn

Comment 12

[Adam Lock]

Attachment: Makefile diff from 0.9.2 branch to invoke manifest generator during build

Comment 13

[Adam Lock]

Attachment: Updated gtkEmbed patch. Contains profile service and a dummy prompt service

Comment 14

[Adam Lock]

I have checked gen_mn.pl into the trunk and attached diffs to this bug for
changes to the makefile to use the script. I have also attached a diff against
gtkEmbed which installs a dummy prompt service as well as including the earlier
profile patch. The dummy prompt service ensures we don't run into issues
associated with posing XUL based prompt messages.

I haven't tried gen_mn.pl on the trunk for a while so I will try that now.

Comment 15

[Chris McAfee]

7/6 - 7/9 patches get https working for me on the trunk.
We still have form focus problems, e.g. you can't login
to bank accounts.  I know what needs to happen for that,
that should be another bug.

Comment 16

[Adam Lock]

r=mcafee

Comment 17

[Chris McAfee]

form focus bug is bug 83201

Comment 18

[Brendan Eich [:brendan]]

Indentation is whacked in InitProfile (tabs, four then two space, etc.) and
InitPromptService (four, then three?!).  Please fix.  Otherwise, I can rubber
stamp this.  Blizzard should bless when he resurfaces, but it seems fine to
check this in first, then get any sr= comments from him that I missed.

Is there a bug on the strange locprovider ownership?  Should there be?

sr=[e-mail] with some indentation uniformity (it'll warm blizzard's
heart).

/be

Comment 19

[Chris McAfee]

fix checked into trunk, with indent fixes for main.cpp that brendan requested ;-)

Comment 20

[Chak Nanga]

Running make in embedding/config on Windows fails with the following error after
the latest checkin. Any ideas?

make-jars -v -d ..\..\dist\WIN32_D.OBJ\bin\chrome -f jar -s ..\..\dist\WIN32_D.O
BJ\Embed\tmpchrome
+++ making chrome D:/mozilla_src/mozilla/embedding/config  => ..\..\dist\WIN32_D
.OBJ\bin\chrome/embed.jar
error: file '..\..\dist\WIN32_D.OBJ\Embed\tmpchrome/help/content/help/help.xul'
doesn't exist at ..\..\config\make-jars.pl line 207, <STDIN> line 2.
NMAKE : fatal error U1077: 'perl.exe' : return code '0x2'
Stop.
NMAKE : fatal error U1077: '"D:\Program Files\Microsoft Visual Studio\VC98\bin\N
MAKE.EXE"' : return code '0x2'
Stop.

Comment 21

[Chris McAfee]

Attachment: 0.9.2 branch patch of trunk checkin, some white space changes (sorry)

Comment 22

[Adam Lock]

Attachment: Updated gen_mn.pl, tolerant of non-existant paths, list moved to top for easy access, problem with $locale substitution sorted out

Comment 23

[Adam Lock]

Chak is seeing problems with the latest embed-jar.mn on Win32 because the help 
module is not built on Win32 by default.

I believe the easiest way to sort these kind of problems is to get gen_mn.pl 
turned on and throw away the static embed-jar.mn. My previous attachment 
cleans the script up somewhat so I will now try and get it working on Win32.

Comment 24

[Adam Lock]

Attachment: Updated gen_mn.pl, works on Win32. Chak can you apply the patch and see if it fixes your problems?

Comment 25

[TVL]

fyi-the 0.9.2 branch patch doesn't include the prompt service files.

Comment 26

[Chris McAfee]

TVL: prompt files are on the trunk.

Branch patch needs updating by now, we didn't quite
get the trunk right to pass all the other architectures
(Win32, IRIX).  Branch patch + trunk prompt files works
for me on a default linux build.

Comment 27

[Chris McAfee]

Attachment: prompt.tar.gz, two trunk files needed for branch patch

Comment 28

[Keyser Sose]

*** Bug 89109 has been marked as a duplicate of this bug. ***

Comment 29

[Chris McAfee]

fix checked in on the trunk, marking fixed.
092 branch people are looking at TestGtkEmbed, please
reopen this bug and resummarize to a branch bug
if needed.

Comment 30

[Michael Dunn]

Clean up verification of dated code change bus