Juan Lang wrote:
Hi Chris,
[../wine-git/dlls/crypt32/rootstore.c:346]: (error) Resource leak: dir
I just sent a patch to fix this. Thanks! --Juan
If the group wants I can run this daily when I run wine test as well.. and post it here.... just let me know
Chris
If the group wants I can run this daily when I run wine test as well.. and post it here.... just let me know
Probably not. Most of the remaining leaks are in the tests or the tools. Since these are short-lived programs, the leaks will get plugged as soon as the programs end anyway. The only exception is the ntdll one, and it only happens when ntdll can't connect to the wineserver, and the process is dying anyway.
Thanks, --Juan
On Tue, Aug 4, 2009 at 2:07 PM, Juan Langjuan.lang@gmail.com wrote:
If the group wants I can run this daily when I run wine test as well.. and post it here.... just let me know
Probably not. Most of the remaining leaks are in the tests or the tools. Since these are short-lived programs, the leaks will get plugged as soon as the programs end anyway. The only exception is the ntdll one, and it only happens when ntdll can't connect to the wineserver, and the process is dying anyway.
Still wouldn't hurt to fix them.
It may not be worth running daily, but perhaps weekly and see if anything new pops up.
On Tue, 4 Aug 2009, Juan Lang wrote:
Still wouldn't hurt to fix them.
Well, Alexandre routinely rejects such "fixes."
It may be worth reconsidering.
Just like if you have hundreds of compilation warnings the useful ones get lost in the sea, if such these tools report too many valid leaks, it will be hard to notice the important ones, or the new leaks.
In any case I see no reason not to fix those in the conformance tests. Yes they may take a few extra milliseconds to run, but they are not performance critical anyway.
2009/8/5 Francois Gouget fgouget@free.fr:
On Tue, 4 Aug 2009, Juan Lang wrote:
Still wouldn't hurt to fix them.
Well, Alexandre routinely rejects such "fixes."
It may be worth reconsidering.
Just like if you have hundreds of compilation warnings the useful ones get lost in the sea, if such these tools report too many valid leaks, it will be hard to notice the important ones, or the new leaks.
grep -v "/tests/"
In any case I see no reason not to fix those in the conformance tests. Yes they may take a few extra milliseconds to run, but they are not performance critical anyway.
-- Francois Gouget fgouget@free.fr http://fgouget.free.fr/ Computers are like airconditioners They stop working properly if you open WINDOWS
Francois Gouget fgouget@free.fr writes:
On Tue, 4 Aug 2009, Juan Lang wrote:
Still wouldn't hurt to fix them.
Well, Alexandre routinely rejects such "fixes."
It may be worth reconsidering.
Just like if you have hundreds of compilation warnings the useful ones get lost in the sea, if such these tools report too many valid leaks, it will be hard to notice the important ones, or the new leaks.
In any case I see no reason not to fix those in the conformance tests. Yes they may take a few extra milliseconds to run, but they are not performance critical anyway.
If there are real leaks in the tests, then they should be fixed, but hacking the code to silence false positives is not a good idea. There are more and more analysis tools out there, we can't add workarounds for every limitation of every tool. False positives should be reported to the tools authors instead.
-
Alexandre Julliard -
Austin English -
Ben Klein -
chris ahrendt -
Francois Gouget -
Juan Lang