I posted a patch back in march which got lost in the pile. http://www.uclibc.org/lists/uclibc/2008-March/019155.html
Thx -Khem On Thu, Nov 13, 2008 at 8:21 PM, Rob Landley <[EMAIL PROTECTED]> wrote: > On Thursday 13 November 2008 18:55:31 Doug Graham wrote: >> On Thu, Nov 13, 2008 at 06:48:41PM -0600, Rob Landley wrote: >> > On Thursday 13 November 2008 13:35:57 Doug Graham wrote: >> > > This is probably the wrong mailing list to report this bug to, but it's >> > > the one I subscribe to, and this bug is relevant to this list, so you >> > > might be interested. >> > > >> > > uClibc's mknod implementation looks like this: >> > >> > Not only a report to the wrong list, but no mention of which version you >> > encountered the problem in (when said project had a new release >> > _yesterday_). >> >> Well, I shouldn't think it would be too hard to check the latest release >> to see if this bug still exists, but you're right, I should have mentioned >> which release I was looking at. It was uClibc-0.9.30 (today's release). > > Ok, worth taking 15 minutes to look at then. > > mknod.c is getting major() and minor() macros out of include/sys/sysmacros.h, > where major() minor() and makedev() are redirected to gnu_dev_*, which is > insane for 3 reasons: > > 1) What does gnu have to do with anything? We're making a linux kernel > syscall. > 2) It's messing around with 64 bit numbers, but kdev_t is a 32 bit number. > (The whole << 32 bit is just bloat that's going to get chopped off, as is the > gratuitous conversion to long long on 32 bit platforms.) > 3) In this case, dev_t and kdev_t should be in exactly the same format, so if > it was working _correctly_ the translation would be a NOP. > > Possibly the attached patch? > > Rob > > _______________________________________________ > uClibc mailing list > [email protected] > http://busybox.net/cgi-bin/mailman/listinfo/uclibc > _______________________________________________ uClibc mailing list [email protected] http://busybox.net/cgi-bin/mailman/listinfo/uclibc
