101949 - Make sure that Qt toolkit's |CreateDeviceContextSpec| returns an error code on failure

Status: VERIFIED FIXED

(Core :: Printing: Output, defect)

Description

[Roland Mainz]

Make sure that Qt-toolkit version of |CreateDeviceContextSpec| returns an error
code on failure. Recently I fixed the error handling in nsDocumentViewer.cpp
incl. the following two lines:
-- snip --
-    factory->CreateDeviceContextSpec(mWindow, devspec, aSilent);
-    if (nsnull != devspec) {
+    rv = factory->CreateDeviceContextSpec(mWindow, devspec, aSilent);
+    if (NS_SUCCEEDED(rv)) {
-- snip --
which means that it isn't "legal" anymore to turn a random value (incl. NS_OK)
on failure as error code. The new code _only_ looks at the error code (and
fires-up an error dialog except for NS_ERROR_ABORT (which indicates that the
print dialog has been canceled... - see bug 101289)) and does not make any
further checks. 
Unix/Linux(GTK+/Xlib) and Windows (bug 101221) have already been fixed (Mac is
bug 101947).

Comment 1

[John C. Griggs]

I think the existing code is OK.  It will only return NS_OK if 
nsComponentManager::CreateInstance manages to create a non-NULL 
nsIDeviceContextSpec instance and if that instance returns NS_OK from it's Init 
method (after being cast to an nsDeviceContextSpecQT, of course.)  I chased the 
code into nsDeviceContextSpecQT::Init and it also looks fine - it returns 
NS_ERROR_FAILURE if there is any problem and NS_OK only if everything works out 
properly.  If CreateDeviceContextSpec encounters any errors creating or Initing 
the DeviceContextSpec, it returns NS_ERROR_FAILURE as well.

Let me know if this needs to be modified.

Comment 2

[Roland Mainz]

Note that my new code fires-up an error dialog for every error you return -
except for NS_ERROR_ABORT (which means the print dialog itself has been canceled
by the user).

Comment 3

[John C. Griggs]

I believe that all of the errors that the code currently returns are actually 
errors - i.e. an error dialog is appropriate.  Right now, printing is not 
implemented in the Qt version of Mozilla, so I will make a note to take care 
with return codes when I get to that work (which will be a while, since I am 
currently working on Mozilla in my spare time, which is very limited at the 
moment).  Are there any paritcular failure modes that you are concerned about 
and where you feel an error dialog is not appropriate?

Comment 4

[Roland Mainz]

> Are there any paritcular failure modes that you are concerned about 
> and where you feel an error dialog is not appropriate?

The only thing is NS_ERROR_ABORT which does not trigger an error dialog (it is
used to indicate that the user has canceled the print session somehow).

Comment 5

[John C. Griggs]

OK.  Since printing is not yet implemented in Qt Mozilla, we don't have to worry 
about cancelled print sessions (yet).  I am marking this fixed.

Comment 6

[sujay]

Roland/John, please verify and mark verified-fixed...thanks!

Comment 7

[Roland Mainz]

VERIFYING per Grigg's comments...