Dave Miner wrote: > Moinak Ghosh wrote: >> Dave Miner wrote: >>> Moinak Ghosh wrote: >>>> Hi, >>>> >>>> I have made some more changes to libtransfer: >>>> >>>> Handle zero-length files to avoid hardlink problems when copying >>>> from hsfs >>>> A separate list is created for all zero length files to avoid >>>> passing >>>> these pathnames to cpio. A separate loop is then used to create >>>> these files onn the harddisk. >>>> >>>> Add couple of flushes to sync buferred file data for ease of debugging >>>> Make cpio use /tmp instead of /var/tmp to avoid filling up the ramdisk >>>> TMPDIR env var is being set to /tmp to force cpio to use swap. >>>> cpio uses >>>> /var/tmp by default and /var/tmp is not mounted on swap. >>>> >>>> The webrevs are at: >>>> >>>> http://jurassic.eng/~mg147109/libtransfer/webrev/ >>>> >>>> I was able to get a successful boot off harddisk without any >>>> sharetab or >>>> utmpx errors after using this library with a DC iso image. >>>> >>> >>> 646: putenv(3c) is quite specific that automatic variables shouldn't >>> be used. >>> >>> 903: seems to me like just using open() would be a wee bit faster; >>> essentially you're doing the same thing as creat(2). >> >> Okay changed, testing shortly. BTW I also realized that the code is >> not preserving the file ownership and perms. So have added storing >> and using the st_mode, st_uid and st_gid values. I will post the new >> changes after testing. >> >>> >>> 904: logging an error if the file isn't created for some reason >>> seems necessary. >> >> That is just a debug message not error handling. I have moved the >> debug >> message to actually print only if it successfully created the file. >> > > I believe you misunderstood the comment: I would like you to log an > error message if the creation fails, because that's an error in the > image.
Ah okay. Regards, Moinak. > > Dave
