Mailing List Archive

Notes on 5.001n
experiences with 5.001n . . .

* My first attempt on each platform is:
./Configure -des
make
make test

* Table below gives a summary, with specific notes following.

* Most of these are "in progress", but i wanted to mail off this much,
partly with a hope of getting a "fixed in my version" or two :-)

* If things settle down, and i have time, i'll start looking into
the core dumps i'm getting.

neilb

Perl 5.001n compile results
===========================

Compiles Passes all Use Just Note
Clean? Testsuite? Defaults?

IRIX 4.0.4 NO YES YES 1

IRIX 5.2 YES YES YES

FreeBSD 2.0.5 - 2

DG/UX 5.4R2.01 - NO 3

BSD/OS 2.0 YES* NO 4

OSF1 V3.0 YES* NO NO 5


---------------------------------------------------------------------------

NOTES

1. IRIX 4.0.4

On the IRIX machine, the architecture name expands to IP12-irix . . .
Kinda surprising. Someone who knows SGIs better might want to comment
on whether irix4 and irix5 are good enough.

One compile warning (apart from the ever-present one in POSIX.c):
toke.c, line 1037: void * and function pointers are not
convertible to each other


2. FreeBSD

Dies at the following point in Configure:

Figuring out the flag used by open() for non-blocking I/O...
Seems like we can use O_NONBLOCK.

Let's see what value errno gets from read() on a O_NONBLOCK file...
./Configure: 5562: Syntax error: ";" unexpected


3. DG/UX

Needs the MakeMaker patch posted separately.
Have to specify 'cc' and NOT 'gcc', else Configure blows up
which is strange, since the hints say you need to use gcc,
and it was me put that in there!
I had to make a small change to Socket.xs -- compile bombed out.
I'll mail this off separately, since i'm not sure on
the validity of the patch.
Currently failing three tests. Working on this one at the moment.


4. BSD/OS 2.0

Needs the MakeMaker patch (posted separately)

Fails two tests:
anydbm core dumps on test 3
ndbm core dumps on test 3


5. OSF1 V3.0

Taking the defaults will try and do dynamic linking, which resulted
in some MakeMaker strangeness, which I'll look into.

Opting for static linking results in a clean compile, and passes all
tests except for socket.t.