Feedback on bug reports

Thomas Kuehne thomas-dloop at kuehne.cn
Sat Mar 17 04:24:46 PDT 2007


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

Bill Baxter schrieb am 2007-03-17:

<snip>

> That's great for a truly open-source project, but the problem with that 
> suggestion for D is that no-one besides Walter, and maybe now Andrei, is 
> really a "member" of this project.  No one besides Walter has any 
> ownership, so nobody really has any more obligation (or authority) to 
> respond to posts than anyone else.  There probably are people here who 
> would step up to the plate and take on such responsibility if they were 
> given some authority, but no-one has such authority now.  Especially if 
> you want to know "why isn't bug 2345 being fixed?".  No-one knows that 
> besides Walter.  So if someone is going to communicate that to you it 
> has to be someone who has Walter's ear.

Walter most certainly isn't the only one capable to fix the compiler.
Ignoring inline assembler (because GDC-0.23 doesn't support it on
AMD64) roughly 9% of DStress'[1] test cases with unexpected results
behave differently under DMD-1.009 i686 and GDC-0.23 AMD64. This
suggest that either DStress contains an overwhelming number of broken
test cases or a large proportion of all observed compiler issues are
due to the shared open source frontend.

Assuming pessimistically that 50% of the misbehaving test cases are
buggy would result in ca. 360 fixable test cases. A single bug usually
causes 1-10 failing test cases. Thus there are at least 36 bugs everyone -
yes this includes you - can fix and submit patches to GDC and/or DMD.

While the communication issue raised by the GPs is valid the question
why bugs aren't fixed seems to be a much more pressing one.

Thomas

[1] http://dstress.kuehne.cn


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

iD8DBQFF+9yELK5blCcjpWoRAujDAJ9WHt/JT1X15dSGUvQ+xOvZaaaz/wCfe/ov
2p4Ei3P1YXVwJPSC3RndGlo=
=gxpo
-----END PGP SIGNATURE-----



More information about the Digitalmars-d mailing list