Bug 91514

Summary: [Qt] feature detection in orwt/nrwt does not work with force_static_libs_as_shared
Product: WebKit Reporter: Balazs Kelemen <kbalazs>
Component: Tools / TestsAssignee: Balazs Kelemen <kbalazs>
Status: RESOLVED FIXED    
Severity: Normal CC: abarth, dpranke, hausmann, ojan, vestbo, webkit.review.bot
Priority: P2 Keywords: LayoutTestFailure, Qt, QtTriaged
Version: 528+ (Nightly build)   
Hardware: Unspecified   
OS: Unspecified   
Attachments:
Description Flags
Patch none

Description Balazs Kelemen 2012-07-17 10:07:25 PDT
The test harnesses analyze the webcore lib with nm to find out what is enabled and what is not. With force_static_libs_as_shared the libQtWebKit.so actually not contains much, we need to do it better.
Comment 1 Balazs Kelemen 2012-07-17 10:25:03 PDT
Created attachment 152780 [details]
Patch
Comment 2 Dirk Pranke 2012-07-17 12:05:46 PDT
Comment on attachment 152780 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=152780&action=review

mostly looks fine ... one question:

> Tools/Scripts/webkitdirs.pm:743
> +        if ($isSearchingForWebCore) {

isn't $isSearchingForWebCore only true if WebCore is in the name? If so, does your s/// actually do anything?
Comment 3 Balazs Kelemen 2012-07-18 02:49:54 PDT
(In reply to comment #2)
> (From update of attachment 152780 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=152780&action=review
> 
> mostly looks fine ... one question:
> 
> > Tools/Scripts/webkitdirs.pm:743
> > +        if ($isSearchingForWebCore) {
> 
> isn't $isSearchingForWebCore only true if WebCore is in the name? If so, does your s/// actually do anything?

Yes, because at the very first we overwrite the parameter:
if (isQt()) {
    $libraryName = "QtWebKit";
because normally libQtWebKit contains every symbol (and it seems like only features.pm use this function anyway). I know it's not really correct, but I have to keep this hacked logic to make it work with a default build and also with force_static_libs_as_shared.
Comment 4 Balazs Kelemen 2012-07-18 02:51:55 PDT
Forgot to mention the motivation of fixing this: I could not run some tests without the patch. For example compositing tets are skipped, because the test harness believes AC is disabled.
Comment 5 Dirk Pranke 2012-07-18 12:44:15 PDT
(In reply to comment #3)
> (In reply to comment #2)
> > (From update of attachment 152780 [details] [details])
> > View in context: https://bugs.webkit.org/attachment.cgi?id=152780&action=review
> > 
> > mostly looks fine ... one question:
> > 
> > > Tools/Scripts/webkitdirs.pm:743
> > > +        if ($isSearchingForWebCore) {
> > 
> > isn't $isSearchingForWebCore only true if WebCore is in the name? If so, does your s/// actually do anything?
> 
> Yes, because at the very first we overwrite the parameter:
> if (isQt()) {
>     $libraryName = "QtWebKit";
> because normally libQtWebKit contains every symbol (and it seems like only features.pm use this function anyway). I know it's not really correct, but I have to keep this hacked logic to make it work with a default build and also with force_static_libs_as_shared.

Oh, right. Okay.
Comment 6 Balazs Kelemen 2012-07-18 14:36:39 PDT
Comment on attachment 152780 [details]
Patch

Clearing flags on attachment: 152780

Committed r123021: <http://trac.webkit.org/changeset/123021>
Comment 7 Balazs Kelemen 2012-07-18 14:36:46 PDT
All reviewed patches have been landed.  Closing bug.