Hi William, I have only nits:
auto_install.c -------------- 431, 458: comment says 'write' - it probably should be 'finalize' 426 * there be no info about partitions for TI, -> 426 * there is no info about partitions for TI, As far as dealing with slice 1 when dedicated to swap device is concerned, your suggestion below sounds reasonable. Thank you, Jan William Schumann wrote: > Jan, > This was put on hold for the November release, but it is high priority > for the SPARC release. > > Jan Damborsky wrote: >> Hi William, >> >> in general the changes look good, I have >> only couple of nits - please see below. >> Also, could you please attach test procedures >> used when validating that new modifications >> work and no regressions were introduced ? >> > Testing partition and slice editing: > -set up AI server > -ran test copy of auto-install and liborchestrator out of /tmp > -created test manifest based on ai_combined_manifest.xml > -copied ai_manifest.default.xml and ai_manifest.rng into /tmp > -command line: LD_DBG_LVL=4 LD_LOAD_LIBRARY=/tmp /tmp/auto-install -p > <my test manifest> > -did several permutations of custom slices and partitions in manifest > > Regression testing, a.k.a. does GUI still work? : > -used RC2 ISO with VirtualBox > -put new builds of liborchestrator.so.1 and libti.so.1 into /tmp > -run as root "crle -e LD_LIBRARY_PATH=/tmp" to add /tmp as a ld search > path > -verify with ldd /usr/lib/gui-install that test shared libraries are used > -as "jack", LS_DBG_LVL=4 pfexec gui-install > -checked /tmp/install_log - call to TI looks fine, swap slice created > when needed > -used format(1M) to check slices and partition > --- saw swap slice in slice 1, slice 0 used rest of available space > -tested with and without predefined Solaris2 partition >> >> auto_td.c >> --------- >> om_write_partition_table(), om_write_vtoc() >> - it seems after those functions were changed, >> they don't carry out real target modifications - >> this happens in do_ti(). Could you please rename >> those functions as the current names are misleading - >> I might recommend something like >> om_prepare_partition_table(), om_prepare_vtoc() > renamed several of these >> >> perform_slim_install.c >> ---------------------- >> >> It seems slim_set_slice_attrs() is no longer used. >> If this is the case, could you please completely >> remove that function from slim_util.c ? >> > Removed, and that left only two functions in slim_util.c - for > preparing TI attributes for slices and partitions, which I moved into > disk_slices.c and disk_parts.c respectively, and deleted slim_util.c. > Changes in slim_util.c pertaining to swap and dump creation were > merged. This brings up a question of what is proper behavior if the > user has other plans for slice 1, but AI decides that swap and dump > should be dedicated to slice 1. Code will allow customization of > slice 1 to take priority over using slice 1 for swap and dump. > > Note that this does not have SPARC-specific code yet. > William >> >> >> William Schumann wrote: >>> http://cr.opensolaris.org/~wmsch/bug-4233/ >>> http://defect.opensolaris.org/bz/show_bug.cgi?id=4233 >>> >>> Calling TI twice for slices and partitions was fixed by using a >>> single calling point for each in perform_slim_install.c for GUI and AI. >>> >>> The actual installation failure turned out to be due to a TI bug in >>> which mktemp(3c) was being called using a pointer to a constant text >>> string, which mktemp() was overwriting. On the second call to TI, >>> the constant text area wasn't valid (did not contain "X"s) and >>> mktemp() failed. >>> _______________________________________________ >>> caiman-discuss mailing list >>> caiman-discuss at opensolaris.org >>> http://mail.opensolaris.org/mailman/listinfo/caiman-discuss >>
