66020 - add CJK/intl support to GtkEmbed

Status: VERIFIED FIXED

(Core :: Internationalization, defect)

Description

[kill this account]

we need someway to support non-latin1 display and non-latin1 applications in 
GtkEmbed.

- need converters to convert CJK, etc. to Unicode

- need ability to set default encoding (ie: a menu item)

- need ability to set override encoding (ie: a menu item)

- need abilit to set the font for a particular language

Comment 1

[kill this account]

missing i18n libs:
libchardet.so required for meta-charset to work
 
libucvcn.so required for Chinese
libucvibm.so required for IBM encodings
libucvja.so required for Japanese
libucvko.so required for Korean
libucvtw.so required for Chinese
libucvtw2.so required for Chinese

Comment 2

[kill this account]

these i18n libs need to be copied/symlinked into the 
mozilla/dist/Embed/components dir.

Comment 3

[Judson Valeski]

is there some way to select the converters we need (at build time? perhaps 
a built flag?) rather than dropping all of them in. gtkEmbed is a lightweight 
example of embedding and if clients use it to do any sizing info, it would be 
good if it didn't include extra libs.

Comment 4

[kill this account]

Attachment: changes to Makefile.in and basebrowser-unix

Comment 5

[kill this account]

Attachment: basebrowserintl-unix - list of intl libs to install

Comment 6

[kill this account]

Attachment: changes to nsEmbedAPI.cpp to start/stop metacharset detection service

Comment 7

[kill this account]

The changes to Makefile.in, basebrowser-unix, and the new file 
basebrowserintl-unix allows a developer to control the installation of the intl 
converter libs.

The change to nsEmbedAPI.cpp starts/stops the metacharset detection service.

Comment 8

[Judson Valeski]

what exactly does the metacharset service do?

Comment 9

[Judson Valeski]

it seems to me that the parser would know when to fire up the meta charset 
stuff; no?

Comment 10

[kill this account]

the windows equivlant of this bug is 
http://bugzilla.mozilla.org/show_bug.cgi?id=66030

Comment 11

[kill this account]

there is a long (not always technical) history to the design.
a per document redesign would be nice.

Comment 12

[Judson Valeski]

I think there may be a better way to init this service. investigating.

Comment 13

[kill this account]

in mozilla the metacharset service gets called at startup:

Here is the stack when the metacharset gets started:
main (nsAppRunner.cpp#1198)
  calls main1 at nsAppRunner.cpp#1272
main1 (nsAppRunner.cpp#811)
  calls nsAppShellService::Initialize at nsAppRunner.cpp#933
nsAppShellService::Initialize (nsAppShellService.cpp#123)
  calls NS_WITH_SERVICE(nsIMetaCharsetService,... at nsAppShellService.cpp#139

Comment 14

[Judson Valeski]

understood. doug and I are suggesting that this model is wrong though. isn't 
there some lower level event that occurs that would be a more appropriate tie 
in?

Comment 15

[kill this account]

Isn't this bug is about:

    "a feature that works in mozilla does not work in embedding"

If we want to request a redesign of how it gets done in mozilla shouldn't we 
open a separate bug?

Comment 16

[Judson Valeski]

bstell's right, this has morphed.

I've opened up seperate bugs for the bigger problem: 
http://bugzilla.mozilla.org/show_bug.cgi?id=67576
http://bugzilla.mozilla.org/show_bug.cgi?id=67578

Comment 17

[Judson Valeski]

I've made 67576 a blocker of this one. the changes in nsEmbedAPI.cpp put build 
time dependencies on low level services. embedAPI is for high level, app 
specific, necessities.

Comment 18

[Teruko Kobayashi]

Changed QA Contact to [e-mail] for now.

Comment 19

[Frank Tang]

why 67576 is blocking this one ? IS this a meta tracking bug ? Can we do 67576
after moz 0.9 and finish other thing in this bug first ?

Comment 20

[Judson Valeski]

67576 blocks this because that needs to be done in order for observer 
registration to occur implicitly. The mods to nsEmbedAPI.cpp are invalid. You 
can't add random components to startup like this. 67576 could be removed as a 
blocker of this, if the charset service used our new (almost checked in) 
startup category. It's up to i18n which route they want to take, either 68720 
blocks this, or 67576 blocks this. 

Comment 21

[Roy Yokoyama]

Jud: Can you explain what the new startup category (2001-03-01 13:07) is?
I am finishing up the blocker 67576; but I am not sure if I am covering 
your new stuff.  Bug 67576 is simply to convert CharDet to use NS_GETMODULE().
Please let us know. :)

Comment 22

[Roy Yokoyama]

Got more info about thenew startup category. See
http://lxr.mozilla.org/seamonkey/search?string=APPSTARTUP_CATEG

Comment 23

[Roy Yokoyama]

Attachment: Use of Category manager. CharDet is defined as "app_startup"

Comment 24

[Roy Yokoyama]

Above patch includes the patch for the blocker 67576.  
I am waiting for /sr= for 67576.  

If you like to see what the patch for this bug, here it is.
(only few changes in nsCharDetModule.cpp and nsMetaCharsetObserver.cpp)
This works great. I tested in winEmbed.  Now it automatically registers
the MetaCharsetDetector and it can display the www.netscape.com/ja correctly.


================ nsCharDetModule.cpp =======================================

+#include "nsIAppStartupNotifier.h"
+#include "nsICategoryManager.h"

+static NS_METHOD nsMetaCharsetObserverRegistrationProc(nsIComponentManager 
*aCompMgr,
+                                          nsIFile *aPath,
+                                          const char *registryLocation,
+                                          const char *componentType,
+                                          const nsModuleComponentInfo *info)
+{
+  nsresult rv;
+  nsCOMPtr<nsICategoryManager> 
+    categoryManager(do_GetService("@mozilla.org/categorymanager;1", &rv));
+  if (NS_SUCCEEDED(rv)) {
+    rv = categoryManager->AddCategoryEntry(APPSTARTUP_CATEGORY, "Meta Charset",
+                        "service," NS_META_CHARSET_CONTRACTID,
+                        PR_TRUE, PR_TRUE,
+                        nsnull);
+  }
+  return rv;
+}

// Component Table
static nsModuleComponentInfo components[] = 
{
 { "Meta Charset", NS_META_CHARSET_CID, 
    NS_META_CHARSET_CONTRACTID, nsMetaCharsetObserverConstructor, 
-    NULL, NULL},
+    nsMetaCharsetObserverRegistrationProc, NULL},
 { "Document Charset Info", NS_DOCUMENTCHARSETINFO_CID, 
    NS_DOCUMENTCHARSETINFO_CONTRACTID, nsDocumentCharsetInfoConstructor, 


======== nsMetaCharsetObserver.cpp ===========================
NS_IMETHODIMP nsMetaCharsetObserver::Observe(nsISupports*, const PRUnichar*, 
const PRUnichar*)
 {
-    return NS_ERROR_NOT_IMPLEMENTED;
+  // "app_startup" category calles nsIObserver::Observe method when app/embed 
enumerates the members
+  return Start();
 }                                                  

Comment 25

[Roy Yokoyama]

Could someone review the code? The patch for this bug only is fine with me. :)
[ie. above changes in nsCharDetModule.cpp and nsMetaCharsetObserver.cpp]
Valeski? Alecf?

Comment 26

[Alec Flett]

thanks for switching to generic modules!
my only request is that you use nsCOMPtrs instead of raw pointers, and remove
uses of ReleaseService() as it's deprecated (and equivalent to NS_RELEASE())

Comment 27

[Roy Yokoyama]

Attachment: New patch: only applicable for this bug. Using nsICategoryManager to start up the MetaCharsetService

Comment 28

[Roy Yokoyama]

alecf: Thank you for your valuable comment.
I've checked in the bug 67576 which includes your suggestions.
(ie. use of nsCOMPtrs instead of raw pointers and remove ReleaseService())
I have attached a new patch to use the category manager.
I am also going to remove the code from nsAppShellService::Initialize().
We don't need the code for nsIMetaCharsetService any longer.

dougt,valeski: can any of you /r=?
alecf: can you /sr=?

Thanks

Comment 29

[Judson Valeski]

looks good to me. r=valeski.

Comment 30

[Roy Yokoyama]

Attachment: Removing MetaCharset from Appshell

Comment 31

[Roy Yokoyama]

Attachment: Removing MetaCharset from Test ViewerApp

Comment 32

[Alec Flett]

this is great - I'm glad to see the junk removed from apprunner.
my only request is that you make sure you're actually getting Observe() on the 
app-startup topic, so that in case someone has registered you for other topics, 
that you aren't called over and over

Comment 33

[Roy Yokoyama]

Attachment: As per Alecf's recommendation: checking the topic in ::Observe()

Comment 34

[Roy Yokoyama]

alecf: Is this good for check in? Can you /sr=?   Thanks

Comment 35

[Alec Flett]

sr=alecf

Comment 36

[Roy Yokoyama]

checked-in.

Comment 37

[Roy Yokoyama]

New checkin in 66030.  (04/16/01 17:07)  See comments.
MetaCharset detector should also be enabled in gtkEmbed.

Comment 38

[Andreas Becker]

Changing QA contact back to teruko.

Comment 39

[Roy Yokoyama]

You have to do the following:
===========Comment from 66030===========
Now the meta charset detector is in place,
it is up to the embedder to include which language he/she
likes to support.  
If he/she would like to support:
-Japanese, copy components/ucvja.dll
-korean, copy components/ucvko.dlland so on.

Comment 40

[Tracy Walker [:tracy]]

this has been fixed since at least march 23.  I have inadvertantly been testing 
the same bits each day since then.

marking fixed

Comment 41

[Teruko Kobayashi]

This was fixed while ago.