William Schumann wrote: > defect: AutomatedInstaller cannot find disks when default manifest is used > > After discussion of the original bug report, it became clear that more > explicit logging would be very helpful to determine the behavior of AI in > evaluating disks for being install candidates. > > During development of these changes, some bugs were found and fixed (below). > > http://defect.opensolaris.org/bz/show_bug.cgi?id=5739 > http://cr.opensolaris.org/~wmsch/bug-5739/ > > Added logging for all code paths, reporting: > - method of evaluation of target disk(manifest criteria, default) > - reason for rejection of target disk candidate (failure to meet any manifest > criteria, disk too small) > > Found bugs: > - not considering manifest element target_device_use_solaris_partition when > determining whether to take default action > - not using partition information for x86 when determining free > space(introduced in new code for bug 4195 > <http://defect.opensolaris.org/bz/show_bug.cgi?id=4195> for SPARC support) > - using slice information not yet available (very minor) > > Unit tested combinations of criteria specifications on x86 & SPARC. >
A few nits, looks pretty good: auto_install.c, 417: I'd modify this message to something like "Use of existing fdisk partition requested" so that it's a little less geeky. auto_parse.c, 268: why not just auto_log_print for this one? auto_td.c, 377: can we just delete this comment? it's documented adequately in om_get_min_size, which is the right place for the issue, and I'm skeptical we would remember to fix this comment when that gets changed. Dave
