On Fri, 2006-12-01 at 21:08 +0000, Al Viro wrote: > On Fri, Dec 01, 2006 at 02:52:11PM -0600, Russell Cattelan wrote: > > code clean up are not without risk and with no regression test suite to > > verify > > that a "cleanup" has not broken something. Cleanups are very much a > > hindrance to stabilization. With no know working points in a code > > history it becomes difficult > > to bisect changes and figure out when bugs were introduced > > Especially when cleanups are mixed in with bug fixes. > > > > Pretty code does not equal correct code. > > No, but convoluted and unreadable code ends up being crappier due > to lack of review. And that's aside of the memory footprint, > likeliness of bugs introduced by code modifications (having in-core > and on-disk data structures with different contents and the same C > type => trouble that won't be caught by compiler), etc.
Nothing makes up for the complete lack of GFS2 testing. reviewed code does not equal correct code either. Honestly tell me what test suite do you run on GFS2? Sure is it possible to make an educated guess that some cleanups will not destabilize the code. Indeed the stuff you have done is quite useful to ensure that endian bugs are being caught by the compiler/sparse. But no amount of "it shouldn't break anything" assertions can replace testing. But there is a large quantity of the 70 or so patches that were sent out were to enable "future" cleanup's. Putting in partial cleanups do nothing core code readability and I many cases is more confusing. Unless you meticulously keep up with the partial cleanups looking at the code is now a jumbled mess of inconsistencies. gfs2 is supposed to be stabilized and use-able for the up coming rhel5 release, not pretty up for somebody to print out and hang on their wall. -- Russell Cattelan <[EMAIL PROTECTED]>
signature.asc
Description: This is a digitally signed message part
