Bug 17691 – REGRESSION: FindSafari doesn't work

• Component: Platform
• Status: RESOLVED FIXED

[Robert Xiao]

I downloaded the very latest nightly (r30768) and found that the FindSafari.exe program doesn't start with this error:

This application has failed to start because the application configuration is incorrect. Reinstalling the application may fix this problem.

I have reinstalled Safari 3.0.4 and redownloaded/unpacked the package, to no avail. The older FindSafari from r30628 works fine, even with the new nightly. Also, the old one is 10.5 KB; the new one is 39.0 KB.

The second issue (cosmetic) is that after starting WebKit with run-nightly-webkit.cmd, the version number as reported by Safari is still 523.15. I know I am running WebKit, though, because it gets a high score (88) on Acid3 (among other things), while regular Safari (when run directly) gets 39.

[Mark Rowe (bdash)]

How are you checking the WebKit version number?  Safari's "About" box displays Safari's version number, not WebKit's, so it will be unaffected by running a nightly build.

[Robert Xiao]

I am using Help->About. I couldn't find out how to do it any other way. Is there an about: page for it?

[Adam Roben (:aroben)]

(In reply to comment #2)
> I am using Help->About. I couldn't find out how to do it any other way. Is
> there an about: page for it?

No, the WebKit version number is not shown anywhere in Safari's UI. It is, however, included in the user agent string, and sites like <http://whatsmyuseragent.com/> make it easy to see the contents of that string.

[Robert Xiao]

OK, so that's not a bug. Thanks for letting me know. The FindSafari thing is still problematic, though.

[Adam Roben (:aroben)]

The r30768 nightly works just fine for me on XP Pro SP 2 (I'm using it right now to write this comment).

Perhaps you should try deleting the %TMP%\WebKitNightly directory and running the nightly again?

[Robert Xiao]

Cleared out my temp directory. Same error, if I use the new FindSafari.

[Robert Xiao]

I should add that it works (but seems to crash on RSS feeds, something I can't seem to reproduce always).

[Adam Roben (:aroben)]

(In reply to comment #7)
> I should add that it works (but seems to crash on RSS feeds, something I can't
> seem to reproduce always).

Does this mean that Safari now launches and you no longer get the error dialog?

[Adam Roben (:aroben)]

(In reply to comment #0)
> I downloaded the very latest nightly (r30768) and found that the FindSafari.exe
> program doesn't start with this error:
> 
> This application has failed to start because the application configuration is
> incorrect. Reinstalling the application may fix this problem.
> 
> The older FindSafari from r30628 works fine, even with the new nightly.
> Also, the old one is 10.5 KB; the new one is 39.0 KB.

It's been suggested that this is all caused by r30394 <http://trac.webkit.org/projects/webkit/changeset/30394>, which caused FindSafari to always link against the debug C runtime. This makes sense to me, though it doesn't explain why r30628 works for you.

Perhaps we should change FindSafari to always link against the release C runtime.
> 
> The second issue (cosmetic) is that after starting WebKit with
> run-nightly-webkit.cmd, the version number as reported by Safari is still
> 523.15. I know I am running WebKit, though, because it gets a high score (88)
> on Acid3 (among other things), while regular Safari (when run directly) gets
> 39.
> 

[Robert Xiao]

Safari only launches (and works, and crashes randomly on RSS...) if the FindSafari.exe file is the one from the r30628 package (posted in the Windows nightlies archive). Simply using the run-nightly-webkit.cmd script with the newer (broken) FindSafari results in the console window appearing momentarily; no dialog error is shown, and Safari doesn't start (as verified by the Process Manager).

If I add a pause statement, the output is:
        1 file(s) copied.
The system cannot execute the specified program.
Setting environment for using Microsoft Visual Studio 2005 x86 tools.

[Robert Blaut]

I can confirm the problem. After executing  a script "run-webkit-nightly.cmd" I get "the system cannot execute the specified program".

[Robert Blaut]

Marked as regression P1 bug.

[Matt Lilek]

(In reply to comment #9)
> (In reply to comment #0)
> > I downloaded the very latest nightly (r30768) and found that the FindSafari.exe
> > program doesn't start with this error:
> > 
> > This application has failed to start because the application configuration is
> > incorrect. Reinstalling the application may fix this problem.
> > 
> > The older FindSafari from r30628 works fine, even with the new nightly.
> > Also, the old one is 10.5 KB; the new one is 39.0 KB.
> 
> It's been suggested that this is all caused by r30394
> <http://trac.webkit.org/projects/webkit/changeset/30394>, which caused
> FindSafari to always link against the debug C runtime. This makes sense to me,
> though it doesn't explain why r30628 works for you.
> 
> Perhaps we should change FindSafari to always link against the release C
> runtime.
>

I just changed it to use our release vsprops instead of the debug ones and the nightly works fine now on my non-dev Vista work box.

[Matt Lilek]

Created attachment 19569 [details]
s/Debug/Release

[Adam Roben (:aroben)]

Comment on attachment 19569 [details]
s/Debug/Release

r=me

[Matt Lilek]

Fix landed in r30864, it will be included in the next nightly.

[Robert Xiao]

FindSafari still doesn't work in r30865, giving the exact same problem as before; the file is 39.0 KB (same size as the not-working r30768 version, but the md5sums are different).

The old version from r30628 still works fine.

[Matt Lilek]

(In reply to comment #17)
> FindSafari still doesn't work in r30865, giving the exact same problem as
> before; the file is 39.0 KB (same size as the not-working r30768 version, but
> the md5sums are different).
> 
> The old version from r30628 still works fine.
> 

It's an issue with the way the nightlies are being built/packaged.

[Robert Xiao]

Alright, I see. Well, I'll continue to use the old FindSafari for now, then. Is there a possibility that using this "outdated" version is resulting in my non-reproducible random crashes, or might that be something else?

[Robert Xiao]

Sorry, I checked the wrong box. It's not fixed.

[Matt Lilek]

(In reply to comment #19)
> Alright, I see. Well, I'll continue to use the old FindSafari for now, then. Is
> there a possibility that using this "outdated" version is resulting in my
> non-reproducible random crashes, or might that be something else?
> 

No - there haven't been any changes to FindSafari is awhile so the "old" version you have will be identical to the "correct" one in ToT.

[Mark Rowe (bdash)]

Grab the latest nightly build (r30868).  It was just fixed.

[Robert Xiao]

I can verify that it is fixed. Thanks!

[Γριφεγ]

No it is not fixed!

I have just downloaded the r34752 nightly.

It still exhibits the same problem on Windows XP!

[Mark Rowe (bdash)]

Given that this bug report was closed *three months ago*, any issue that you are seeing should be filed as a separate bug report as it's likely to be an unrelated issue.

[Γριφεγ]

Oops... Not having Safari installed results in the same symptoms as this bug.

The WebKit nightly download site should tell the rest of us that it needs a Safari installation, I think...