[Bug 11184] New: Performance regression in variable substitutions.

http://bugs.winehq.org/show_bug.cgi?id=11184

Summary: Performance regression in variable substitutions.
           Product: Wine
           Version: 0.9.53.
          Platform: PC-x86-64
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: -unknown
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

Created an attachment (id=10240)
 --> (http://bugs.winehq.org/attachment.cgi?id=10240)
Benchmark with Wine 0.9.52 and 0.9.53 doing the same thing.

After installing wine version 0.9.53, I discovered a big performance hit in
some scripts i used before:

I removed  ~/.wine/ but it did not changed anything.

[Bug 11184] Performance regression in variable substitutions.

http://bugs.winehq.org/show_bug.cgi?id=11184

--- Comment #1 from Antonio   2008-01-14 04:04:46 ---
Compiled from source using:
gcc version 4.1.0 (Gentoo 4.1.0-r1 p1.8)
On:
Linux 2.6.22 Gentoo-sources.

[Bug 11184] Performance regression in variable substitutions.

http://bugs.winehq.org/show_bug.cgi?id=11184

--- Comment #2 from Roderick Colenbrander   2008-01-14 05:04:38 ---
As of Wine 0.9.53 wineboot is started at wine startup to launch services (and I
think the drive detection will be a service soon too, so it is needed). Most
likely this is what is adding some additional delay. If no trace of Wine is
around when starting winepath, this likely happens. Try to see what happens if
you have e.g. winemine running when you launch your script. Likely it defaults
to the old behavior.

[Bug 11184] Performance regression in variable substitutions.

http://bugs.winehq.org/show_bug.cgi?id=11184

--- Comment #3 from Antonio   2008-01-14 06:22:37 ---
(In reply to comment #2)
> As of Wine 0.9.53 wineboot is started at wine startup to launch services (and I
> think the drive detection will be a service soon too, so it is needed). Most
> likely this is what is adding some additional delay. If no trace of Wine is
> around when starting winepath, this likely happens.

Thanks for answering, but please, pay more attention to what happens.
Additional 3 seconds delay happens only when the "winepath" executable is
invoked in a variable substitution.
`winepath` or $(winepath) takes 3 seconds more than winepath to complete.
> Try to see what happens if
> you have e.g. winemine running when you launch your script.

Zool ~ # winemine &
[1] 31797
Zool ~ # time winepath ; sleep 5 ; time echo $(winepath)

real    0m0.014s
user    0m0.007s
sys     0m0.007s

real    0m0.013s
user    0m0.003s
sys     0m0.007s
> Likely it defaults
> to the old behavior.

Exactly! but i'm still asking why $() and `` adds an additional 3 seconds
delay, and if there is a workaround apart from launching another wine app in
background.

[Bug 11184] Performance regression in variable substitutions.

http://bugs.winehq.org/show_bug.cgi?id=11184

Vitaliy Margolen  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |trivial
           Keywords|                            |regression

--- Comment #4 from Vitaliy Margolen   2008-01-14 08:54:08 ---
Trivial problem at best. You are welcome to perform regression testing to find
the culprit of the problem.

[Bug 11184] Performance regression in variable substitutions.

http://bugs.winehq.org/show_bug.cgi?id=11184

Vitaliy Margolen  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |WONTFIX

--- Comment #5 from Vitaliy Margolen   2008-01-14 09:01:08 ---
And you need to learn how shell works. 'winepath' does not wait for the child
processes to exit, $(winepath) does. That's your extra 3 seconds for starting
wineboot.

Nothing to fix here.

[Bug 11184] Performance regression in variable substitutions.

http://bugs.winehq.org/show_bug.cgi?id=11184

Alexandre Julliard  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |UNCONFIRMED
         Resolution|WONTFIX                     |

--- Comment #6 from Alexandre Julliard   2008-01-14 09:14:25 ---
Obviously they both wait for the process to exit, the difference is that in the
$(winepath) case there's also a pipe that doesn't get closed until the server
exits. It's clearly a bug.

[Bug 11184] Performance regression in variable substitutions.

http://bugs.winehq.org/show_bug.cgi?id=11184

Alexandre Julliard  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |FIXED

--- Comment #7 from Alexandre Julliard   2008-01-14 09:44:59 ---
Should be fixed now.

[Bug 11184] Performance regression in variable substitutions.

http://bugs.winehq.org/show_bug.cgi?id=11184

--- Comment #8 from Antonio   2008-01-14 11:47:34 ---
(In reply to comment #7)
> Should be fixed now.

Thank you.

[Bug 11184] Performance regression in variable substitutions.

http://bugs.winehq.org/show_bug.cgi?id=11184

Antonio  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #9 from Antonio   2008-01-26 06:22:15 ---
Confirmed to be Fixed in 0.9.54.