> > I am progress of doing the same for dagrab (expect a follow-up to PR > > 57227 soon). > > There is no need for extra-patches. > #ifdef CDIOCREADAUDIO seems to be sufficient enough.
The version in GNATs is obsoleted, but my update did not make it there, although it appeared in freebsd-ports-bugs@ (I used the follow-up email-link in the GNATs web interface). I will prepare and test a version that used pread() like yours this night. It will be here then: http://home.leo.org/~barner/freebsd/dagrab-pread.patch > Is using CDRIOCSETBLOCKSIZE necessary as well? > If yes it shouold be done in enable_cdda callback I think. I don't now. I don't know how to interpret Soren's comments in http://freebsd.rambler.ru/bsdmail/freebsd-current_2003/msg15989.html regarding our situation: > The right way of doing audio graps has been to set the wanted > blocksize with CDRIOCSETBLOCKSIZE (not needed if all tracks on > the CD is of the same size), then just read from the device. > Ioctl's was newer meant to be used to read/write data, its also > faster to use the R/W path... Can we assue that all tracks on an audio CD are of the same (block) size? What about those mixed-mode CDs (sorry, I don't know the specifications). > Moreover, is there a reason we should use CDIOCREADAUDIO on RELENG_4? > Please test attached patch. works for me on RELENG_4 and should work > on -CURRENT as well. Yep, it works fine for me on 4.9-STABLE and 5.1-CURRENT (and complies with Soren's posting). I think the best solution would be to patch all applications to use this approach. The only question is whether we need CDRIOCSETBLOCKSIZE. Simon
signature.asc
Description: Digital signature
