>> Lines 101-109 - As I indicated in an earlier review, I don't >> believe this is necessary. Both Nevada and the Solaris 10 >> patch gate do large pages automatically (or so-called out of >> the box) and so including these options is unnecessary. >> However, I've cc'ed Bart Smaalders who is an expert in this >> area who can suggest whether or not it makes sense to include >> this. > > I disagree... the comment explicitly cites the benchmarks (mhhh... I > hoped the current comment was enougth... should I add the whole > benchmark code and all the results as comment there (which raises the > question if there is a size limit for comment sections in Makefiles...) > ? =:-) ). And neither on my Ultra5 (B48, B51) or our university machines
Please no - there's is of course no reason to add such verbiage to the comments I am curious when (against which build) did you do your benchmarks and what sort of workload they consistent of? >> Lines 70, 80, 331-332, 343-344, 384-385, 404-405, various other >> lines in wordexp() - It appears there's some sort of mismerge >> with Roger's >> >> PSARC 2006/659 fork extensions >> 6497356 fork extensions > > Yes, but AFAIK April merged the (current) ksh88 version back (in her > SCCS tree). For the ksh93 version I've talked with Roger Faulkner... his > new version uses |posix_spawn()|&co. (which is IMO a very good idea) but > it's very late now to port these parts to the ksh93 version of > |libc::wordexp()| ... I only stumbled over the new ksh88 version of > |libc::wordexp()| at the end of May and getting the new version > created&&_tested_ will take some time (creating a new patch is no > problem... getting it tested will be another xx@@@!!-story) ... and > pushing a new version without a good testing coverage by all OpenSolaris > distributions for such a very risky part is IMO not a good idea > (remember we have this alternative |libc::wordexp()| version in the tree > because otherwise SMF may blow-up and the OpenSolaris distributions had > problems to deal with the issue... and that's why I am now dragging this > pain with me...). But I'm talking about the code that is *not* under #if WORDEXP_KSH93. Unless I'm misreading the webrev, the *existing* wordexp() was mismerged with Roger's putback. >> usr/src/pkgdefs/SUNWarc/prototype_com >> usr/src/pkgdefs/SUNWarc/prototype_i386 >> usr/src/pkgdefs/SUNWarc/prototype_sparc >> >> I know what's happened to libcmd as part of this project but >> why are you no longer delivering the lint libraries via these >> three files (especially since you're updating llib-lcmd >> itself?) > > The Solaris libcmd API was moved to libc and therefore llib-lc takes > over the duties for the |def*()|-API and the ksh93 parts of libcmd are > currently not a public API, therefore we don't deliver libcmd.so and > llib-lcmd ... ;-( So then why are you updating llib-lcmd at all? Why not just remove it? Some potential future hope that it will be made public? dsc
