Re: bitmap loading related reiserfs changes in 2.6.17-mm1 are broken

[Hans Reiser <reiser@namesys.com>]

Andrew Morton wrote:

>It would really help if Chris or one of the namesys developers could take
>the time to review and test these patches closely, please.
>  
>
Chris's code also has QA problems.  Because I don't control their
salaries, they both just basically ignore the QA process that the rest
of Namesys uses, and have exactly the results one would expect.  I no
longer encourage non-Namesys companies to pay our developers directly 
--- because of this experience.

That QA process is: everyone must get their patch reviewed and tested by
a second developer before sending it in.  Elena is not qualified to
review the code, but she is able to run a standard suite of tests (one
of which is gcc....) in addition to whatever special test is required
for the patch.   I review the design aspects of the patches, but leave
to others to find most of the coding errors.

Everyone once in a while someone strays from it, and I complain
privately about it.  The percentage of time that it is a mistake to
stray from it is remarkably high.....

Unfortunately, there is both real user demand for the
on-demand-bitmap-loading, and a proper qa of it is non-trivial.  I
propose that vs look at it next week.  vs, please ack.  Chris, if you
are available, please take a look as well.

Jeff has done a lot of good work in optimizing V3 bitmap related code,
and I would like to thank him for that.  It did a lot for our
performance.  Chris's work on the journaling code was also very
important and I am very grateful for it.  A better QA methodology could
have dramatically reduced the number of bugs in that code though.

Hans