"I. Szczesniak" wrote: > On 6/2/07, Roland Mainz <roland.mainz at nrubsig.org> wrote: > > Grumpf... no... do you want to see my list of "major" items still open > > in ast-ksh.2007-04-18 (which is the basis for this putback (disclaimer: > > None of these bugs are "critical" and ksh93 passes the testsuite and it > > capable of replacing /usr/bin/ksh in it's current version without > > problems, too)) ? There is still _lots_ of work which needs to be > > done... ;-( > > Could post this list, please?
Ok... note that the list apply to the current state of the ksh93-integration prototype005 tree which is based on ast-ksh.2007-04-18 (e.g. we use this version for the first ksh93-integration putback and the following putback will either putback ksh93s+_final or jump directly into the ksh93t-_alpha series (depending on testing and how many of the bugs below can be killed with the updates)): -- snip -- - memory usage is extremely high (my fault... I didn't realise the problem (due lack of a machine which has more than 512MB main memory) caused by an array used for job control which scales with the maximum number of processes - in the worst case 384k may go down the river of tears) - builtin "fds" command doesn't support IPv6 ([f]) - builtin ksh93 manual page (e.g. $ ksh93 --man #) does not list /etc/ksh.kshrc ([f]) - exiting a subshell does not restore the previous locale settings via |setlocale()| ([w]) - "mkdir" builtin does not set the sticky bit ([w], [f]) - binary variables are not working for multibyte locales ([w], [f]) - the internal constants for MAXLINE/MAXWINDOW are too small, resulting in problems with very large terminal windows (e.g. you only hit this with something like a Xinerama setup which spans over multiple screens) ([f]) - libast uses |select()| instead of the preferred (and faster) |poll()| - ksh93's handling of SIGRT is slightly incorrect (it assumes fixed signal values) - |libast::fopen()| does not support the "F" flag - Adding elements to a normal array doesn't "scale", e.g. it becomes slower as more elements are added ([w], [f]) -- snip -- [w]=workaround available at shell script level [f]=fix available in newer ast-ksh release Some of these issues are already fixed in ast-ksh.2007-05-15, for others are workarounds available and others can be simply ignored for normal usage. The only "real" issue which really hurts (me (and likely Moniak, too)) is the memory usage - this is going to be addressed for the ksh93t development series (via fixing the job control array thing and David had more ideas like allocating input/etc. buffers only on demand) ... ---- Bye, Roland -- __ . . __ (o.\ \/ /.o) roland.mainz at nrubsig.org \__\/\/__/ MPEG specialist, C&&JAVA&&Sun&&Unix programmer /O /==\ O\ TEL +49 641 7950090 (;O/ \/ \O;)
