• Don Clugston (13/13) Mar 08 2006 It would be nice if, in the summary page, the 'hotspots' listing
• Thomas Kuehne (29/41) Mar 09 2006 -----BEGIN PGP SIGNED MESSAGE-----

Don Clugston <dac nospam.com.au> writes:

It would be nice if, in the summary page, the 'hotspots' listing 
indicated whether the failures are independent of the command-line 
options, or specific to some of them (ie, only show up in 'torture' 
test). In the boxes where it says "XPASS", "FAIL", etc, it could display 
"XPASS+", "FAIL+", "ERROR+" instead, if there were some command-line 
options where it passed.

Most cases with a + are likely to be Linux-specific, so this would make 
dstress more indicative for Windows DMD.

You could even split the them up that way in the summary, this would 
show that the big change between dmd 0.144 and 0.145 was that most of 
ERROR+ became PASS, whereas ERROR probably didn't change much. This 
would make comparisons between gdc and dmd more precise.

-Don.

Thomas Kuehne <thomas-dloop kuehne.cn> writes:

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Don Clugston schrieb am 2006-03-08:
 It would be nice if, in the summary page, the 'hotspots' listing 
 indicated whether the failures are independent of the command-line 
 options, or specific to some of them (ie, only show up in 'torture' 
 test). In the boxes where it says "XPASS", "FAIL", etc, it could display 
 "XPASS+", "FAIL+", "ERROR+" instead, if there were some command-line 
 options where it passed.

 Most cases with a + are likely to be Linux-specific, so this would make 
 dstress more indicative for Windows DMD.

If you are interrested in Windows, how about running DStress on Windows?
Doing so shouln't be difficult for the vast majority of test cases.
(exception: those in http://dstress.kuehne.cn/complex)

1) install a Subversion client

2) checkout http://dstress.kuehne.cn with Subversion

3) compile dstress.c (might require some adaptation for Windows)

4) for all files in "compile" that don't contain ".svn" in the path:
.\dstress torture-compile compile/<source> 2>> log.txt >>&2

5) for all files in "run" that don't contain ".svn" in the path:
.\dstress torture-run run/<source> 2>> log.txt >>&2

6) for all files in "nocompile" that don't contain ".svn" in the path:
.\dstress torture-nocompile nocompile/<source> 2>> log.txt >>&2

7) for all files in "norun" that don't contain ".svn" in the path:
.\dstress torture-norun norun/<source> 2>> log.txt >>&2

8) please send me the compressed log.txt, so that I can put the results
online

Step 4 to 7 will result in several hours hard work for you CPU.
The compressed log file should be about 1.1M for DMD-0.149.

 You could even split the them up that way in the summary, this would 
 show that the big change between dmd 0.144 and 0.145 was that most of 
 ERROR+ became PASS, whereas ERROR probably didn't change much. This 
 would make comparisons between gdc and dmd more precise.

I've got some ideas how to improve the summary generation, but haven't
got enough time to implement them.

Thomas

-----BEGIN PGP SIGNATURE-----

iD8DBQFEEIIW3w+/yD4P9tIRAo5oAKCdN6oo1EyG5thHJwSsjklxtGtZ6gCdEXql
xEMzQG4zeiF6YyNeFB4eAvs=
=Vuy2
-----END PGP SIGNATURE-----