Hi, > So I've downloaded and compiled 1.3.4 versions of libburn, > libisoburn and libisofs. After that I've tried brasero again and it failed, > but with different error.
What error messages exactly ? It would be no surprise if it fails at different block address or with different error messages, as long as the messages indicate the same failure of ioctl(SG_IO). (See below about commits 5180 + 5187.) Even a single success would be not enough. It must work multiple times in a row, in order to be significant. > Seems to me that libburn/libisoburn is the main > cause of problems. It is nearest to the problem, at least, because it sits between Brasero and the kernel's ioctl(SG_IO). If you can point me to a version that reliably succeeds and one that reliably fails, then there is hope to find the decisive difference. libburn-1.3.2 was released Aug 7 2013 libburn-1.3.4 on Dec 12 2013 Current is http://files.libburnia-project.org/releases/libburn-1.4.0.tar.gz It should work flawlessly with applications which expect older versions since about 0.4.0. All older release tarballs can be found at: http://libburnia-project.org/wiki/OutdatedLibburnReleases libburn has two levels, where differences may have occured: - The system adapter libburn/sg-linux.c which handles all Linux-specific aspects. Among them SG_IO for transport of SCSI/MMC commands and replies. Its history can be inspected via http://libburnia-project.org/log/libburn/trunk/libburn/sg-linux.c No good suspect to see there: Commits 5180 + 5187 promise better error messages in 1.3.4 in comparison to 1.3.2. Commits 5218, 5225, 5335 are supposed to be unrelated to SCSI transport. Commit 5368 was used to diagnose problems with SATA on an ARM board. (USB worked flawlessly on the same machine, SATA produced hick-ups and echoed command relies. Sometimes it worked for a single burn run. Mostly not. The user finally settled with USB.) - The higher functions of libburn could change the wear pattern of controller and drive by causing SCSI commands beeing sent in different sequences. http://libburnia-project.org/log/libburn/trunk/libburn It seems futile to inspect all changsets between 5132 (when 1.3.2 was freshly out and development step 1.3.3 started) and 5426 (after which current release 1.4.0 went out). One would rather have to compare SCSI logs of versions in order to detect decisive differences. In libburn-1.X.Y directory (after ./configre && make, to get libburn compiled and cdrskin/compile_cdrskin.sh to get a staticaly linked cdrskin, which needs no "make install"): cdrskin/cdrskin -V dev=/dev/sr0 blank=as_needed -eject image.iso \ 2>&1 | tee -i /tmp/cdrskin-1.X.Y.log A successful run will produce a WRITE(10) message for every 32 KB. So the log files might become quite large. Have a nice day :) Thomas -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org