home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.lang.perl
- Path: sparky!uunet!stanford.edu!snorkelwacker.mit.edu!bloom-picayune.mit.edu!athena.mit.edu!fabbott
- From: fabbott@athena.mit.edu (Freeland K Abbott III)
- Subject: Help: SCO_3 aliasing?!
- Message-ID: <FABBOTT.92Aug25182403@e40-008-12.mit.edu>
- Sender: news@athena.mit.edu (News system)
- Nntp-Posting-Host: e40-008-12.mit.edu
- Organization: Massachusetts Institute of Technology
- Distribution: comp
- Date: Tue, 25 Aug 1992 22:24:13 GMT
- Lines: 45
-
-
- I'm attempting to compile perl 4.019---no, I'm attempting to use it;
- it compiled just fine. I took hints for i386 and sco_3, and used
- perl's malloc; the rest was taken from the hints, including "undef"
- for $i_varargs. Without perl's malloc, lots of tests failed; with it,
- only lib/big test 61 fails. But I can't find a better result, using
- SCO 3.2 rev4.0....
-
- However, I did find one particularly interesting behavior: here's the
- code:
-
- local(*x,*y);
- $x = 3;
- print("x=$x, y=$y\n");
- $y = 4;
- print("x=$x, y=$y\n");
- $x[0] = "hi";
- $y[0] = "bye";
- print ("$x[0]\n");
-
- produces this output:
- x=3, y=3
- x=4, y=4
- bye
-
- Huh?! Somehow x and y are aliased to be identical...
-
- Anyway, what I need is a Clue. This is straight 4.019, which we have
- working on lots of other machines: it's not perl, it's something in
- the configuration with which I compiled. I'm startled that with such
- a basic problem so many of the tests pass... also, we've got a (much
- more complicated) program which, left to run for a while, causes a
- kernel panic: we need to figure these out, and I've already tried all
- the easy permutations in Configure.
-
- Incidentally, if we declare variables in separate local statements
- (i.e. local(*x); local(*y)... rather nasty in the case of
- local(..)=@_;), we get "correct" (non-aliased) results. The complex
- case still dies, but not in a kernel panic (at least not from just one
- run)... I'm still looking for that one, but need help with the simple
- case first.
- --
- Freeland K. Abbott fabbott@athena.mit.edu
- 104 Madison Ave. now working for Cambridge Technology Group
- Arlington, MA 02174 "And I thought 14-hr days would be over now"
-