Hi,
http://www.winehq.org/site/winapi_stats
Is incorrect as it shows Total 19640 100% 100% would be nice but its incorect!
Tom
On Wed, 31 Mar 2004, Tom wrote:
Hi,
http://www.winehq.org/site/winapi_stats
Is incorrect as it shows Total 19640 100% 100% would be nice but its incorect!
I just noticed that there are some extra dlls to filter out:
ttydrv winealsa.drv winearts.drv wineaudioio.drv wined3d winedos winejack.drv winemp3.acm winenas.drv wineoss.drv wineps wineps16.drv
I guess the .dll extension is implicit. It would be nicer with it but that's just my personal preference.
Also, x11drv was supposed to be filtered out already and yet it is listed, and with 4 stub at that! But if I check today's x11drv.spec file there are no stubs in x11drv so something is very wrong...
Similarly winearts.drv is shown as having 2 stubs out of supposedly 5 APIs and yet it's spec file in its entirety is:
@ stdcall DriverProc(long long long long long) ARTS_DriverProc @ stdcall wodMessage(long long long long long) ARTS_wodMessage @ stdcall widMessage(long long long long long) ARTS_widMessage
Only 3 APIs and no stub here!
Also apparently the 16bit dlls are listed before the 32bit dlls. But the separation is not clear so it's confusing. Personally I'd prefer a plain alphabetical list that mixes 16 and 32 bit dlls. Maybe add '(16bit)' to those dlls that are 16 bit.
As I see you are the only one checking my patches please check http://hatky.myz.info/wine/site/winapi_stats and see if it's ok and it there are any more changes needed
On Fri, 2 Apr 2004, hatky wrote:
As I see you are the only one checking my patches
Well, I know it can be tricky to get the statistics just right. My stats were not 100% correct the first time around either (and they may still not be).
please check http://hatky.myz.info/wine/site/winapi_stats and see if it's ok and it there are any more changes needed
Did you find what caused the API count discrepency for winearts.drv and others?
I checked a bigger dll: user32 and got slightly different results. I believe this is because your script worked on a slightly older version of Wine.
In any case, here is how I checked so you can verify a few other dlls. Count various categories of lines:
* wc -l user32.spec 744
* egrep '^#' user32.spec | wc -l 43
* egrep '^ *$' user32.spec | wc -l 5
* egrep '^@ stdcall' user32.spec | wc -l 630
* egrep '^@ cdecl' user32.spec | wc -l 21
* egrep '^@ varargs' user32.spec | wc -l 2
* egrep '^@ stub' user32.spec | wc -l 43
* egrep '^@' user32.spec |wc -l 696
Now let's do some checks:
* 696=630+21+2+43 This tells us that we did not miss any API type. Good.
* 696=744-43-5 This tells us that we took into account all the lines.
* Then check 696 against the 'API Count' on the statistics page. You have 692 which is consistent with version 1.84 of user32.spec.
* And check 43 against the 'Real Stubs' on the statistics page. You have 40 which is also consistent with version 1.84 of user32.spec.
So user32 seems to be ok. crtdll and msvcrt would be nice dlls to check: they have extern APIs that might be miscounted (they also used to have forward APIs which caused me some trouble in the past).
On Thursday 01 April 2004 22:40, Francois Gouget wrote:
Did you find what caused the API count discrepency for winearts.drv and others?
You said remove them...................
I checked a bigger dll: user32 and got slightly different results. I believe this is because your script worked on a slightly older version of Wine.
It is true I ran api_extract a few weeks ago and do not regenrate it each time, the main site will only regenerate it ever version that comes out, it was noted someplace....
In any case, here is how I checked so you can verify a few other dlls. Count various categories of lines:
All of numbers are take from the perl script winapi_extract with the fallowing line running it
./tools/winapi/winapi_extract --pseudo-stub-statistics --no-verbose --no-progress > ../winapi_stats.txt
My work is only formatin this text file to an HTML page, you can run it, then my php script and it there are missmaches plz tell me If the txt file comming out is wrong, check that perl script to see what's wrong and either submit a patch or post to winedevel...
On Sat, 3 Apr 2004, hatky wrote:
On Thursday 01 April 2004 22:40, Francois Gouget wrote:
Did you find what caused the API count discrepency for winearts.drv and others?
You said remove them...................
Yes but it would still be interesting to know why the statistics where wrong rather than just sweep the problem under the rug.
In your latest patch you filter out more dlls but you also duplicate some entries like 'winenas.drv'. Maybe the dlls were already correctly filtered out and the only problem was that the page I looked at was way out of date...
Yes but it would still be interesting to know why the statistics where wrong rather than just sweep the problem under the rug.
As I told you, you can run the command
./tools/winapi/winapi_extract --pseudo-stub-statistics --no-verbose --no-progress > ../winapi_stats.txt
and check the file (it is humanly readable), then either 1. Send a patch for that perl script your self 2. Open a bug in bugs.winehq.org 3. Report back here
Remeber, this is not paid work, I do what I like, I like sites, and php, I do not know perl and have not wish to learn at this time. I know what your saying, but I have no intrest in checking it, sorry.
In your latest patch you filter out more dlls but you also duplicate
Fixed and resent, 10x...
-
Francois Gouget -
hatky -
Tom