On Mon, Feb 28, 2011 at 08:22:13PM -0500, Phillip Susi wrote: > I assume you didn't mean to take this off list? Adding the list back to Cc. >
oops, yes, this should have gone to the list. > On 02/28/2011 06:02 PM, Brian C. Lane wrote: > >What I meant was that the macro, as it exists now, doesn't do anything > >with the action. It looks like that was changed with commit > >562e0007840f06f475b43bb81ffe81b238b627b5 which notes that PED_ASSERT > >shouldn't be used if you don't want it to exit. > > So the argument is being phased out? Then rather than silently > ignore it, I think a review should be done and the second argument > either dropped, or the calling code be rewritten to not use > PED_ASSERT. Having the vestigial argument there only leads to > confusion and bugs. I think both should be done. Remove the PED_ASSERT as your patch does, and remove the 2nd argument from all remaining uses of it. > > >I don't think they're garbage. fdisk works with them just fine. I think > >there's a bug in the way parted tries to determine the CHS in some edge > >cases. I don't know what these are, and I don't think it actually > >matters -- it should be just fine to return 0 in these cases. > > Does fdisk try to guess geometry like this? I didn't think it does, > or if it does, then it falls back to sane defaults when the guess > fails. I didn't check the math in any specific metadata to see if > it was a corner case that the code does not handle correctly, but > whether there is a bug or not, you really can't rely on the CHS > values and so shouldn't be bailing out if they turn out to be bogus. > sfdisk is where the fdisk chs code lives, it looks to me like it is guessing, but the code isn't as well commented as parted's. -- Brian C. Lane | Anaconda Team | IRC: bcl #anaconda | Port Orchard, WA (PST8PDT)
pgpXrjFP5tcJX.pgp
Description: PGP signature
_______________________________________________ parted-devel mailing list [email protected] http://lists.alioth.debian.org/mailman/listinfo/parted-devel
