Re: [Pnet-developers] Re: [DotGNU]Portable.net on AIX 5+

[Peter Colson]

On 02/12/2004, at 1:49 PM, Rhys Weatherley wrote:

> On Thursday 02 December 2004 10:04 am, Peter Colson wrote:
>
> > > That stuff comes from the third-party libgc package, so we don't
> > > really know
> > > much about how it works across AIX versions.
> >
> > Yes that didn't click when I typed it in. I picked those 2 references,
> > but saw others (i.e., support/regex.c and cscc/cpp/ansidecl.h) so
> > thought that there had been some AIX work done in pnet.
>
> Those are also third party code, imported into pnet.  The core code of 
> pnet is
> written to be as generic as possible.  Greping for "#ifdef AIX" is not 
> a good
> way to see if code will run on AIX.  I know that the following code 
> probably
> works on AIX although personally I've never tried it:
>
>     #include <stdio.h>
>     int main(int argc, char *argv[])
>     {
>         printf("Hello World!\n");
>         return 0;
>     }
>
> By your logic, because that code doesn't contain "works on AIX 5L 5.2" 
> in the
> README, it couldn't possibly work!  Or that I must regular testing on 
> every
> AIX version known to man on the off-chance that IBM decides to rewrite 
> AIX to
> disobey Posix.

Comparing "hello world" to pnet is disingenuous. I've seen a fair bit 
of the pnet code and in addition to it, libgc, libffi, etc., etc. I 
think it's valid to try and ascertain ahead of time if it's been seen 
to work in a particular environment (as has been reported on the 
dotgnu.org) site. Blake's post on the older(?) developers list re. 
AIX5+ raised concerns for me.

> Portable.NET is written to be as portable as possible.  There are some 
> ugly
> things like libgc and libffi that are unavoidably non-portable, but we 
> try to
> keep them isolated.
>
> If you want to know if pnet will run on your particular AIX flavour, 
> how about
> you try running it?  You're asking me for an answer to a question that 
> only
> you have the facilities to answer!
>
> > > I suppose this is a long-winded way of saying that in the absence of
> > > an AIX
> > > volunteer, you just became it. :-)
> >
> > OK, but I'm still trying to determine if pnet has/hasn't been tried on
> > AIX previously? If it has maybe the particular version used has been
> > forgotten?
>
> People have posted on the list in the past (search the archives) that 
> they
> managed to get it going on their particular brand of AIX.  But since 
> no one
> has yet committed to test pnet version to version, it is possible that 
> things
> have been missed.  Feel free to submit a patch.

I searched the pnet-developers archive (the most active one it seems) 
days ago and got "ht://Dig error. htsearch detected an error...etc" 
which I reported to address@hidden I've scanned every thread 
title from March '03 to Dec '04 and there is no reference to AIX (apart 
from my qu's). There is a Nov post on the developers archive from Blake 
Fisher about make problems on AIX 5+ which has gone unanswered. When I 
search on that archive his post comes up along with a bunch of 
"one-star" entries related to hpux and libffi which then can't be 
opened. So unless you have any knowledge of archive entries elsewhere 
that report AIX success (which is what I'm trying to find) then I would 
have to assume it doesn't work on AIX.

Are there any other archives to search???

> > If it hasn't the "AIX" reference on //www.dotgnu.org probably should 
> > be
> > removed.
>
> Then I would have to remove basically everything from the list except
> Linux/i386 and Windows.  And then have to deal with people claiming 
> that it
> doesn't work on anything, when it fact it does work on a wide variety 
> of
> platforms.

I find this hard to believe - you'll claim it works on a platform 
because you "think it should"! That seems very back-to-front to me. You 
may be able to claim that it "should" work on most Linux-based systems 
because that's where it was primarily developed I assume, but to claim 
it works on non-Linux systems sight-unseen is a bit of a stretch. Was 
it claimed to work under Windows because it "should have worked" or 
because it was seen to work? The latter I would have thought, given 
Windows differences to the run-of-the-mill Linux platform. Correct me 
if I'm wrong.

I was asked earlier whether pnet was running under OS/400 (which 
attempts to be Posix-compliant as far as I am aware) and because I 
couldn't see it work I wasn't about to claim that it "should" work....

> We are not going to remove a platform merely because we lack testing
> facilities.  Rather, we should encourage people to correct whatever 
> problems
> come up as time goes on.  All we can say is "we've seen it working on
> platforms A, B, and C in the past".  In the absence of an actual bug 
> report,
> we should not assume that it is broken.

That was my initial question - what AIX platforms has it been seen to 
work on? So far I can't see any evidence of any AIX platforms in the 
archives.

> I'm still waiting on a bug report describing the problem that started 
> this
> thread.  "It doesn't work" is insufficient.

I haven't raised a bug report, I'm asking for prior experience of pnet 
on the platform. I'm still waiting to see if it's ever approached a 
workable state on AIX before I invest more time and effort (and 
finances) investing in an AIX box on the, possibly limited, chance we 
can get pnet working on AIX and transport those binaries to PASE under 
OS/400.


Regards,
Peter Colson.