>- Two scripts (from opensource projects) were found which simply failed >because they were using the the ksh93 builtin commands instead those >from /usr/bin/. This is a Solaris-specific issue as those scripts fail >because they depend on Solaris-specific features of the commands in >/usr/bin/ while the ksh93 builtin commands are a POSIX implementation. >This is clearly a portabilty issue in the script, however we fixed the >problem in the meantime in ksh93.
I think that not blaming ksh93 for this failure is wrong in principle; people are free to write Solaris specific scripts; if such scripts break because of incomplete builtins, then I can only find fault in ksh, not in the script. Shell builtins have always been a wart in shell implementation; there should be only one implementation of each command. If I understand correctly, then we are moving to a common implementation of the builtin commands and the Solaris commands. I shell needs that type of integration with the source OS or it should forego any builtins. >- We found another script which checked whether this is running on >Solaris ksh via examining the ksh version and then the script runs a >Solaris-specific codepath. >Basically in this case there is no real solution for this kind of >problem as the authors (Hi Irek! :-) ) added a broken check (and I >request that noone comes up with the idea to add an option to "fake" the >shell version to make that compatible to the old Solaris ksh - that will >be contraproductive). However the authors have promised to fix that >ASAP. Yeah; that is really buggy. Casper
