Jarkko wrote: !That is: ! ! cd t ! ln -s ../perl ./perl ! cd .. ! !> test or extension build that complains should complete successfully. !> Unfortunately not everyone who builds on z/OS can reproduce the !> problem and it has been troublesome to figure out why the Makefile is ! !My money would on the 'make' if this were a betting issue: since it's !the Makefile that is supposed to get that symlink created, and the !Makefile is the same (well, assuming 'sh' which creates the Makefile !works the same always...), and manually creating the symlink works and !resolves the problem, I have to suspect make.
Unfortunately IIRC both users of IBM make and users of GNU make reported not seeing the difficulty. I am not completely sure which make was used by folks who see the problem. Note that both make utilities are customizable via rules files and the IBM make would not work at all if it's rule file is not copied from /samples to /etc. Since many folks on linux, *BSD, Solaris, etc. are also using GNU make I could only speculate at this point: perhaps there is an ordering problem due to the EBCDIC character set for the port of GNU make to z/OS that renders it capable of deferring or ignoring the symlink pre-test rule (???). For reasons that I'd have a hard time explaining we had not hit the bug in the port of GNU make before. More speculation: if the problem is occuring with users of IBM make then perhaps the folks who had successful builds with that utility (which includes you Jarkko IIRC) had a later patched version of IBM make. Perhaps we should be recommending a particulat make utility PTF for folks who've reported this problem? Peter Prymmer
