1052395 - Improve ipc debug output for unhandled nonqueued Windows messages

Status: RESOLVED FIXED

(Core :: DOM: Content Processes, defect)

Description

[Jim Mathies [:jimm]]

I'm working on a patch that improves this output for better debugging.

Comment 1

[Jim Mathies [:jimm]]

[14384] ###!!! ASSERTION: Received "nonqueued" ui message "WM_PAINT" during a synchronous IPC message for window 0x2C0538 ("MozillaTaskbarPreviewClass"), sending it to DefWindowProc instead of the normal window procedure.
[14384] ###!!! ASSERTION: Received "nonqueued" ui message (0x11B) during a synchronous IPC message for window 0x2C0538 ("MozillaTaskbarPreviewClass"), sending it to DefWindowProc instead of the normal window procedure.
[14384] ###!!! ASSERTION: Received "nonqueued" WM_USER message (0x400) during a synchronous IPC message for window 0x2C0538 ("MozillaTaskbarPreviewClass"), sending it to DefWindowProc instead of the normal window procedure.
[14384] ###!!! ASSERTION: Received "nonqueued" WM_APP message (0x8000) during a synchronous IPC message for window 0x2C0538 ("MozillaTaskbarPreviewClass"), sending it to DefWindowProc instead of the normal window procedure.
[14384] ###!!! ASSERTION: Received "nonqueued" registered windows message (0xC2F6) during a synchronous IPC message for window 0x2C0538 ("MozillaTaskbarPreviewClass"), sending it to DefWindowProc instead of the normal window procedure.
[14384] ###!!! ASSERTION: Received "nonqueued" system message (0x10000) during a synchronous IPC message for window 0x2C0538 ("MozillaTaskbarPreviewClass"), sending it to DefWindowProc instead of the normal window procedure.

Comment 2

[Jim Mathies [:jimm]]

Attachment: patch v.1

Comment 3

[Jim Mathies [:jimm]]

Attachment: patch v.1

- without my test output calls DumpNeuteredMessage

Comment 4

[Ben Turner (not reading bugmail, use the needinfo flag!)]

Comment on attachment 8471562 [details] [diff] [review]
patch v.1

Review of attachment 8471562 [details] [diff] [review]:
-----------------------------------------------------------------

This looks pretty good to me!

::: widget/windows/WinUtils.cpp
@@ +60,5 @@
>  namespace mozilla {
>  namespace widget {
>  
> +EventMsgInfo gAllEvents[] = {
> +  {"WM_NULL",                             0x0000},

I think these would be better with a macro, something like:

  EventMsgInfo gAllEvents[] = {
    #define ENTRY(_msg) { #_msg, _msg }

    ENTRY(WM_NULL),
    ENTRY(WM_CREATE),
    // ...

    #undef ENTRY
  }

That way you don't have to worry about typos in the name or value.

::: widget/windows/WinUtils.h
@@ +76,5 @@
> +typedef struct {
> +  const char * mStr;
> +  UINT         mId;
> +} EventMsgInfo;
> +extern EventMsgInfo gAllEvents[];

Seems like this should be DEBUG-only right?

Comment 5

[Jim Mathies [:jimm]]

> I think these would be better with a macro, something like:
> 
>   EventMsgInfo gAllEvents[] = {
>     #define ENTRY(_msg) { #_msg, _msg }
> 
>     ENTRY(WM_NULL),
>     ENTRY(WM_CREATE),
>     // ...
> 
>     #undef ENTRY
>   }
> 
> That way you don't have to worry about typos in the name or value.

great idea, updated.

> ::: widget/windows/WinUtils.h
> @@ +76,5 @@
> > +typedef struct {
> > +  const char * mStr;
> > +  UINT         mId;
> > +} EventMsgInfo;
> > +extern EventMsgInfo gAllEvents[];
> 
> Seems like this should be DEBUG-only right?

Hmm, not really, but it could be. The code in nsWindowDbg that references this works in both debug and release. Not sure if we need to clean that up, the linker should get rid of this table if its not in use.

Comment 6

[Jim Mathies [:jimm]]

Attachment: patch v.2

v1 try push - https://tbpl.mozilla.org/?tree=Try&showall=0&rev=0e5318389b69

debug mochitest-3 has some of these new warnings in it.

v2 try push - https://tbpl.mozilla.org/?tree=Try&showall=0&rev=d7dbb7912c60

Comment 7

[Jim Mathies [:jimm]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/b3c4148afdd4

Comment 8

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/b3c4148afdd4