Peter Memishian wrote: > > > Why so many complains for one simple bug? > > It is not one simple bug, it is the latest in a sequence of bugs related > to this change. There is a case to be made that some of these bugs could > have been caught prior to integration given more testing, such as 6790805 > and 6793120. There were also more subtle bugs (such as 6805584) that I > doubt we would've caught. There are others that have not yet been > root-caused, such as 6800517.
Erm... what is http://bugs.opensolaris.org/view_bug.do?bug_id=6800517 ? > In the end, it is all time lost :-/ I disagree that this is "lost time". The bug we're hitting here ([1]) is something which can be traced many months back which caused noise in script benchmarking and we never found the root cause by quickly looking at it. _Now_ we know what really caused this mess&&slime. []1=Techncially this is a bug in ksh93. The alias.sh and alias.c wrapper contain the use of "~(modifer)pattern"-style shell pattern where "modifer" is "Elr" which means that the following pattern is an extended regular expression pattern with left and right anchors. The bug in this case is that the shell sees the leading '~' character and then does some kind of "proactive" |getpwnam()| call which result isn't used in this case - and that's causing the name service call and the resulting "forks&&torches"-style discussion... ;-( We have two solutions in the work: 1. Temporary remove the use of "~(modifer)pattern"-style shell pattern in alias.c 2. Permanently fix the shell that it only uses |getpwnam()| when it's really needed and not just because something starts with a '~' character. > > Maybe wrong decision, maybe good (less code duplicity). > > There are lots of ways to share code. FWIW, in the case of sleep, there > wasn't a whole lot to share -- the venerable old binary was all of 94 > lines of source code, including comments. Yes, and it wasn't bug-free. > Sure, it doesn't have all sorts > of bells and whistles, but it also just worked and got out of your way, > which is what a simple, core utility should do. > > To be clear, I am not opposed to the idea of sharing code -- far from it. > I am also not opposed to the idea of using the AST codebase to subsume > grotty commands that are complex or have conformance/maintenance issues. > However, sleep was simple and harmless, and last required a change in > 1996, hence my opposition. The point of using "sleep"&&"sum" first (before going to "cut"/"paste"/etc.) was to make sure the concept works in the "wilderness". As I was trying to point out in tools-discuss@ we really can't predict all bugs or problems and some issues with the wrapper caught us off-guard (like the process name issue... it's obious when you spend some time to think about it but none of us who had the knees deeply in the code realised that this may be a problem). But we're trying to learn and deploy fixes quickly. ---- Bye, Roland -- __ . . __ (o.\ \/ /.o) roland.mainz at nrubsig.org \__\/\/__/ MPEG specialist, C&&JAVA&&Sun&&Unix programmer /O /==\ O\ TEL +49 641 3992797 (;O/ \/ \O;)
