74938 - acrobat fails on plugin registration on Linux

Status: VERIFIED FIXED

(Core Graveyard :: Plug-ins, defect, P3)

Description

[shrirang khanzode]

Linux only (mac got fixed recently)

Steps:
1 Put the nppdf.so file for acrobat plugin inside mozilla plugins folder
2 Go to the url above
3 Observe that the File Save dialog pops up.. (if you do not have acrobat as a 
  helper app configured)
4 typing 'about:plugins' does not list the plugin...

expected: acrobat should appear embedded inside the browser window just like on 
other platforms..

Comment 1

[Peter Lubczynski]

Since this never has worked, I think this is due to the gtk switch over and 
Adobe needs to update their plugin. 

Andrei, thouhts?

If so, then Adobe needs to be evanglized.

Comment 2

[av (gone)]

That's a possibility. But we need to make sure that we at least try to load the 
plug-in.

Comment 3

[yiming chung]

see the same prolem on Solaris. The plugin loading fails during the
initializtion process. 

Immediately after the plugin(nppdf.so) is loaded, nppdf.so's NP_Initialize is
called (see the code below). Note that the call to PR_FindSymbol returns a
function pointer (not null) but the subsequent call to the returned function
fails. The function call, 
 pfnInitialize(&(ns4xPlugin::CALLBACKS),&callbacks),  returns a NPError of 4,
NPERR_MODULE_LOAD_FAILED_REEOR.

///////////////
ns4xPlugin::CreatePllugin
    NP_PLUGINUNIXINIT pfnInitialize =
        (NP_PLUGINUNIXINIT)PR_FindSymbol(aLibrary, "NP_Initialize");

    if (pfnInitialize == NULL)
        return NS_ERROR_UNEXPECTED; // XXX Right error?

    if (pfnInitialize(&(ns4xPlugin::CALLBACKS),&callbacks) != NS_OK){
-->>		return NS_ERROR_UNEXPECTED;
    }
//////////////////////////////////////////////////////////

Comment 4

[av (gone)]

So the library gets loaded and evev an entry point is retreived. Does not look 
like missing linked libraries. Serge, do you have any ideas what could happen 
here?

Comment 5

[Peter Lubczynski]

Changing summary and re-assign to Serge.

Comment 6

[serge (gone)]

av, shrir
did unix mozilla ever handles pdf files using acrobat plugin?
as I recall it does handle it using helper apps, am I right?

Comment 7

[shrirang khanzode]

yes, it always worked as a helper app, never embedded as  a plugin inside the 
browser.

Comment 8

[Vesa Halttunen]

Might this be similar to bug 56464 (RealPlayer plugin only works if Mozilla is
compiled using gcc 2.95.2)?

Comment 9

[Arjen Hommersom]

using a build of gcc 2.95 doesn't solve the problem in this case. 

Comment 10

[cls]

Figures. No sooner do I say that I can't get acrobat to work in 4.x (in bug
84427), it starts working (albeit on a different box).   It still doesn't work
in mozilla and I think it may be due to nppdf.so requiring symbols not found in
the mozilla tree.  Here's the list of undefined symbols referenced by nppdf.so 
(nm --dynamic nppdf.so) but not found in libc, libdl, libX11 or libXt:

CreateQueue
DeQueue
DestroyQueue
EnQueue
NPN_GetURLNotify
CreateQueue
DeQueue
DestroyQueue
EnQueue
NPN_GetURLNotify

List generated using:
 nm --dynamic /usr/lib/netscape/plugins/nppdf.so | grep -v ^0 | sed 's|^        
||' | grep ^U | sed 's|^U ||' > ack 
# That's 9 spaces in the first sed

cat >findlibs<<EOF
#!/bin/sh

checklist="/lib/libc.so.6 /lib/libdl.so.2 /usr/X11R6/lib/libX11.so
/usr/X11R6/lib/libXt.so.6"

for l in $checklist; do
    rm -f found-$l
done
rm -f found-not
for s in `cat $1`; do
    found=0
    for l in $checklist; do
        nm $l | grep -c $s >/dev/null 2>&1
        if [ $? = 0 ]; then
            echo $s >> found-`basename $l`
            found=1
            break;
        fi
    done
    if [ $found = 0 ]; then
        echo $s >> found-not
    fi
done
EOF

# Then
sh -x findlibs ack

Comment 11

[Cameron Smith]

I have got the same problem, using the Acrobat 4.05 plugin.  It works fine in
Netscape 4.77 (by simply dropping nppdf.so in the netscape plugins directory),
but not in mozilla build 2001061308.

Comment 12

[serge (gone)]

Attachment: initial patch proposal

Comment 13

[serge (gone)]

small notice:
to run acrobat reader on linux you have to 
setenv PATH <wherever acrobat install dir>/bin:$PATH, or  equivalent for your
shell, then try 'acroread' it should start stand alone reader.

Comment 14

[av (gone)]

Wow! Acrobat is now working in the browser window! Thanks for the great effort,
serge.

Although, there are still some problems with the patch like it doesn't behave
correctly on resizing the browser window, I think it worth checking into the
trunk. The changes in GetPluginInfo code are likely to help with the long
standing problem we have with starting Mozilla when audio device is active,
because it will not try to initialize plugins on start up. 

r=av

Comment 15

[av (gone)]

Changing milestone as we have progress here.

Comment 16

[Chris Waterson]

The original design was to have ns4xPlugin wrap _all_ interfaces to 4.x plugins.
I don't know if that is relevant anymore, maybe av or beard can comment. If it
_is_ relevant, than promoting the logic to nsPluginsDirUnix.cpp is not right; we
ought to continue to hide 4.x plugin logic from the brave new world. I'd suggest
either altering ns4xPlugin::Create() to simply return a wrapped, but
un-initialized plugin, or adding a static method to ns4xPlugin that can be used
to extract the MIME type given a pLibrary or something.

The changes in ns4xPlugin look fine to me.

Comment 17

[av (gone)]

This is good point. Although we decided to deprecate the new api (the Mozilla
Plugin API as it has been known) and encourage people to stick with old api with
some new additions to it for scriptability and access to the browser services
(now we are promoting name NPAPI), I think we are still going to keep the
Mozilla API in place as some vendors already invested in it and released
commercial products based on it. So, the suggestion looks reasonble. I think
something like ns4xPlugin::GetPluginInfo will be good to have.

Comment 18

[serge (gone)]

Chris, thank you for comments.
I completely agree it's better to keep all legacy (4.x) interfaces in one place, 
but those stuff is platform specific, and ns4xPlugin.cpp has a lot of #ifdef  
it's getting difficult to read, maybe it's time to break it by platforms.
Sure, I'm going to implement something like static ns4xPlugin::GetPluginInfo for 
unix.

Comment 19

[av (gone)]

Nominating for the branch. This patch is going to fix a long standing and very
annoying issue known as "hang on start up when audio device is active".

Comment 20

[serge (gone)]

Attachment: ns4xPlugin::Create() returns not initialized plugin object

Comment 21

[av (gone)]

Well... This sort of parameter overloading may seem confusing, I would prefer a
clear flag, but not insisting. r=av.

Comment 22

[serge (gone)]

I agree, but I have strong feeling we have to split ns4xPlugin.cpp by platforms 
or toolkits, and I do not want to add any platform specific methods into 
that now.

Comment 23

[Marek Z. Jeziorek]

nsbranch+ per pdt triage
please try to fix it within next couple of weeks

Comment 24

[Chris Waterson]

Attachment 49297 [details] [diff] looks fine; sr=waterson. Let's get a bug on file to break
ns4xPlugin up by platform as serge suggests. I agree that would be a good idea.

Comment 25

[serge (gone)]

can I get PDT+ for check into the branch?

Comment 26

[Philip Tellis]

I was able to view pdf documents in 0.9.3 using the nppdf.so plugin from
acrobat, but after switching to 0.9.4, it doesn't work anymore.

Comment 27

[serge (gone)]

Do you mean you were able to run acrobat on unix as helper apps, not in browser 
window, am I right?

Comment 28

[Jaime Rodriguez, Jr.]

Adding correctness Status Whiteboard, correct/expected behavior does not occur.

Comment 29

[serge (gone)]

checked into the trunk
modules/plugin/base/src/ns4xPlugin.cpp,v  <--  ns4xPlugin.cpp
new revision: 1.71; previous revision: 1.70
modules/plugin/base/src/nsPluginsDirUnix.cpp,v  <--  nsPluginsDirUnix.cpp
new revision: 1.14; previous revision: 1.13

Comment 30

[serge (gone)]

resolve as FIXED

Comment 31

[Peter Lubczynski]

I don't think this was checked into the 0.9.4 branch. Looks like it still needs
a PDT+ approval. Reopening...

Comment 32

[serge (gone)]

that's right, I did not check it into the 094 branch yet

Comment 33

[shrirang khanzode]

hm..on today's trunk on linux 0924, my acroreader is not getting launched as a 
plugin. I get an error in the console.

500 Cannot exec acroread or acroexch errorno=2
Acrobat Plugin-Could not launch acrobat

Comment 34

[serge (gone)]

see Comments From [e-mail] 2001-09-06 18:44

Comment 35

[serge (gone)]

warning messages like:
"Warning: Actions not found: addBookmark, viewBookmark, copy, undefined-key,
find, findAgain, history, loadImages, openURL, mailNew, new, openFile, print,
exit, reload, saveAs, paste, delete, cut, undo, historyItem, back, forward,
abort, PageUp, PageDown
Warning: Actions not found: ManagerGadgetNextTabGroup,
ManagerGadgetPrevTabGroup, DrawingAreaInput, addBookmark, viewBookmark, copy,
undefined-key, find, findAgain, history, loadImages, openURL, mailNew, new,
openFile, print, exit, reload, saveAs, paste, delete, cut, undo, historyItem,
back, forward, abort, PageUp, PageDown
Warning: No action proc named "ManagerGadgetArm" is registered for widget "form"
Warning: No action proc named "ManagerGadgetArm" is registered for widget "form"
Warning: No action proc named "ManagerGadgetArm" is registered for widget "form"
...
are coming from acrobat because it's trying to install 4.x translations which is
irrelevant for 6.x browser.

Comment 36

[Jussi-Pekka Mantere]

Shrirang, have you been able to test this in the trunk changing the PATH
variable? Need to confirm before getting PDT+ for this.

Comment 37

[shrirang khanzode]

finally..yes. Awesome, Serge ! I have tested this fix on the linux 0924 trunk 
and it looks very good. I see the warning messages that serge mentioned above. 
Pls grant pdt+

Comment 38

[Jaime Rodriguez, Jr.]

Pls update the Patch Status for both patches. Check it into the branch - PDT+

Comment 39

[serge (gone)]

cannot update attachments status, got error "You are not authorized to edit
attachments" :(

Comment 40

[serge (gone)]

checked into the MOZILLA_0_9_4_BRANCH
new revision: 1.10.2.1; previous revision: 1.10
new revision: 1.67.2.2; previous revision: 1.67.2.1

Comment 41

[shrirang khanzode]

looks good on today's linux branch (0927). VERIFIED.