search

Perl / perl5

🐪 The Perl programming language
https://dev.perl.org/perl5/
Other
1.98k stars 559 forks source link

RE: 5.6.0 ready for prime-time? #1730

Closed p5pRT closed 20 years ago

p5pRT commented 24 years ago

Migrated from rt.perl.org#2981 (status was 'resolved')

Searchable as RT2981$

p5pRT commented 24 years ago

From pete@petes-place.com

Old marketing surveys from 96-97 showed that companies were moving from Novell and Unix to NT because they thought they had no choice. Since that time\, that choice has been a painful and expensive one. If we can't straighten out microsoft for them\, we can at least give them high quality tools.

Bizarre. You have a truely strange view of Perl's place in the universe. How exactly is Perl supposed to help people trying to use NT in place of Unix? Why should we want to?

[cut]

I imagine you can guess what a large portion of perl5-porters thinks of your kind of "advocacy" by now.

I will not enter into this "My OS is better than your OS" senseless and childish debate with you. It has no place in the now heterogeneous and highly eclectic perl community. The porters have spoken very clearly that this anti-Win32 bias\, though obviously present in some members\, is not the norm among the members of this forum​: and I do believe them. Therefore\, don't pretend to defend them by representing your personal\, infinitessimally trivial biases as those of the perl5-porters group.

p5pRT commented 24 years ago

From @samtregar

On Sun\, 2 Apr 2000\, David Grove wrote​:

Bizarre. You have a truely strange view of Perl's place in the universe. How exactly is Perl supposed to help people trying to use NT in place of Unix? Why should we want to?

[cut]

I imagine you can guess what a large portion of perl5-porters thinks of your kind of "advocacy" by now.

I will not enter into this "My OS is better than your OS" senseless and childish debate with you. It has no place in the now heterogeneous and highly eclectic perl community. The porters have spoken very clearly that this anti-Win32 bias\, though obviously present in some members\, is not the norm among the members of this forum​: and I do believe them. Therefore\, don't pretend to defend them by representing your personal\, infinitessimally trivial biases as those of the perl5-porters group.

What a convenient "[cut]". You managed to pull the lower paragraph entirely out of context. Your lack of integrity is appalling.  
The people on this mailing list feel many different ways about the OS-debate. Nothing of that enters into my opinion of your "advocacy"\, nor do I imagine it influences many others here. What decides me against you is your refusal to put forth evidence to back up your slanderous claims.

Your kind of "advocacy" is something we can do without.

-sam

p5pRT commented 24 years ago

From [Unknown Contact. See original ticket]

What a convenient "[cut]". You managed to pull the lower paragraph entirely out of context. Your lack of integrity is appalling.

The two were related and nothing worth noting was in the middle. I don't believe that it was out of context.

The people on this mailing list feel many different ways about the OS-debate. Nothing of that enters into my opinion of your "advocacy"\, nor do I imagine it influences many others here. What decides me against you is your refusal to put forth evidence to back up your slanderous claims.

What do you want me to do? Mirror the perl5-porters mailing list to you? You're obviously not reading it. Other people are giving the evidence\, I'm just commenting on them and pulling the comments together. Other than that\, I've only made one claim on my own\, that Perl 5.6 is out of compatibility with major CPAN modules. Evidence is simple​:

1. Take a Win32 machine (I wouldn't risk my linux systems on this right now). Follow the instructions\, configure the make file\, nmake. Ok\, fine\, compiles.

2. Go to a module. perl Makefile.PL\, nmake ...lots of compiler messages... then boom\, sorry\, the module you have compiled is out of service at this time.

3. Repeat with other modules. Some go in\, but the most important ones fail. I tried in total 20 modules\, some went in\, others failed with the same errors. Win32 Perl is not Win32 Perl without LibWin32. Tk didn't go through (no\, Nick\, that's not a slap)\, I don't remember the others.

Now\, from other posts (oh\, is it a surprise that I'm not the only one?)\, I'm gathering that POLLUTE is a potential way around this. When setting up my makefile\, I basically skipped this section. This may have been my error\, I'll find out soon enough. The makefile had this big\, horrid-sounding message that if you turn this on all kinds of bad things can happen... so I left it off and went on to the next configuration item. Even if\, get that\, even if "perl Makefile.PL POLLUTE=1" solves the problem\, 5.6.0 is still out of sync with CPAN. If a watch runs out of sync with the other watches around it\, you take it to a watchmaker (unless it's a cheap watch). Why? BECAUSE IT'S BROKEN!

Duuuuh.

Maybe there's a way around it\, maybe there isn't. I have yet to try out Nick's fix for the problem. If it doesn't\, then it's _really_ broken. If it does\, then perhaps these symbol changes should have had a patchlevel to catch up. Or maybe in the makefile\, rather than a barely descriptive horror story\, something like "a significant number of CPAN modules will not compile without this\, this\, and this." Maybe. Maybe maybe maybe...

I'll see when I have time to trudge through it again.

It doesn't make any difference at this point. What does make a difference is whether we learned from it or not. You (we) can't put out a release that breaks a significant number of libraries... at the very least not without stern and obvious warnings about them. This is virtually something that needs a big red flashing light. A few little mentions don't cut it.

Your kind of "advocacy" is something we can do without.

We\, as in unix-perl bigots? Again\, I don't believe that the porters appreciate being put into this generalization. They have specifically told me so.

We\, as in the members of the unix-perl community who doesn't want Win32 perl to advance in any way shape or form? I prefer linux too. I'm just not arrogant enough to deny Win32 users the respect they deserve as members of our eclectic community\, though rank newbie many may be.

We\, and yes I believe I speak for the majority of this forum\, would very much like to close this thread and move on to fixing this problem\, or figuring out how to deal with it. This is going to be very confusing to users.

p5pRT commented 24 years ago

From @ask

On Sun\, 2 Apr 2000\, David Grove wrote​:

[...]

left it off and went on to the next configuration item. Even if\, get that\, even if "perl Makefile.PL POLLUTE=1" solves the problem\, 5.6.0 is still out of sync with CPAN.

That I don't see your name at the "current testers" page as tester of platforms Windows like at http​://testers.cpan.org/ surely must be a mistake\, right?

Your efforts would be far more valuable there than they seem to be "here".

A lot of module authors don't have access to a Windows environment\, but I have yet to see any who was not more than happy to get patches improving compatibility with other platforms\, including Windows.

If you really care then you would go and help them.

- ask

p5pRT commented 24 years ago

From [Unknown Contact. See original ticket]

left it off and went on to the next configuration item. Even if\, get that\, even if "perl Makefile.PL POLLUTE=1" solves the problem\, 5.6.0 is still out of sync with CPAN.

That I don't see your name at the "current testers" page as tester of platforms Windows like at http​://testers.cpan.org/ surely must be a mistake\, right?

No. I wasn't aware of it. I am now and subscribed\, and even somewhat excited about it. Permission to email you some questions about this group in private email? The page isn't very explanatory.

Your efforts would be far more valuable there than they seem to be "here".

I'd very likely be more polite there. Even so\, I'll still haunt this group (p5p) from time to time when I see somebody being stepped on.

A lot of module authors don't have access to a Windows environment\, but I have yet to see any who was not more than happy to get patches improving compatibility with other platforms\, including Windows.

Patches I can do. Win98\, NTS4SP5\, SuSE 6.2\, Win2K\, BeOS\, TurboLinux 4\, and likely FreeBSD or whatever ends up on my spare partition in my Pavillion. (I can't get the damn drive out! Might as well use it for something.)

If you really care then you would go and help them.

Just remember one thing\, however. I work 13-18 hours a day on perl already. But I'll pitch in wherever I can. It actually fits right now because getting stable modules on Win32 is a primary goal of CM/PM and CMCD. Maybe I can even get my boss to fund a minimal amount of my time there. He's pretty open\, as long as he makes money in the long run. ;-))

Actually he's been wonderful. Without him there would be no CM 1.0 or CMCD 1.0!

p5pRT commented 24 years ago

From [Unknown Contact. See original ticket]

David Grove \pete@​petes\-place\.com writes​:

Tk didn't go through (no\, Nick\, that's not a slap)\, I don't remember the others.

Then please report it to \nick@​ing\-simmons\.net\, \ptk@​lists\.stanford\.edu or even (as you a blaming it on perl) here. But do it as a bug report (with Tk version\, perl -V output and copy/pasted nmake/compiler messages) so I/we can fix the thing\, not as diatribe.

I repeat - Tk builds and runs for me\, and also obviously for ActiveState as they are distributing a PPM​:

http​://www.ActiveState.com/packages/zips/Perl5.6/Tk.zip

p5pRT commented 24 years ago

From @ask

On Mon\, 3 Apr 2000\, David Grove wrote​:

[...]

That I don't see your name at the "current testers" page as tester of platforms Windows like at http​://testers.cpan.org/ surely must be a mistake\, right?

No. I wasn't aware of it. I am now and subscribed\, and even somewhat excited about it. Permission to email you some questions about this group in private email?

I am actually somewhat out of touch with that project\, but the cpan-testers mailinglist is usually also used for "meta" discussions etc\, so go ahead and ask there. It's excellent if you can start testing the modules on Win32 as I believe that has been a somewhat missed part of the effort in the past.

[...]

A lot of module authors don't have access to a Windows environment\, but I have yet to see any who was not more than happy to get patches improving compatibility with other platforms\, including Windows.

Patches I can do. Win98\, NTS4SP5\, SuSE 6.2\, Win2K\, BeOS\, TurboLinux 4\, and likely FreeBSD or whatever ends up on my spare partition in my Pavillion. (I can't get the damn drive out! Might as well use it for something.)

Excellent. Go ahead and get started. ;-)

- ask

p5pRT commented 24 years ago

From @doughera88

On Sun\, 2 Apr 2000\, David Grove wrote​:

I've only made one claim on my own\, that Perl 5.6 is out of compatibility with major CPAN modules. Evidence is simple​:

1. Take a Win32 machine (I wouldn't risk my linux systems on this right now). Follow the instructions\, configure the make file\, nmake. Ok\, fine\,   ^^^^^^^^^^^^^^^^^^^^^^^^

Please note that following the instructions includes reading loud obvious warnings and workarounds about the POLLUTE issue. This was not an accident\, it was not a mistake\, it was a deliberate desicision. If you failed to read the instructions\, there's little else we can do.

If we had PERL_POLLUTE=1 as the default\, then some modules would break on some systems (e.g. conflicts with system headers or conflicts with other #defines in those modules or supporting libraries) and that breakage would be very hard to fix in a general way. If we had PERL_POLLUTE=undef as the default then some other modules would break on some other systems due to mis-matches in symbol names\, but at least we could document an easy quick fix. So we did.

There was no obvious perfect solution. None. Every solution we considered could cause problems. I would be delighted if you could prove me wrong by providing a perfect solution. Absolutely delighted. Patches welcome.

So what to do? Since you have apparently ignored this information the first time I presented it\, I will not waste my time repeating it\, but we built in both binary compatibility by default (so many folks don't have to rebuild anything) and source compatibility by two different optional routes. We did this about 16 months ago so CPAN authors might have time to catch up and adjust.

What else could we possibly have done?

Yet despite all that hard work and countless hours of volunteer effort put in by others\, you insist on being extremely critical. This makes me sad.

breaks a significant number of libraries... at the very least not without stern and obvious warnings about them. This is virtually something that needs a big red flashing light. A few little mentions don't cut it.

Please suggest a specific patch to the INSTALL file to change make the warning sufficiently big.

  Andy Dougherty doughera@​lafayette.edu   Dept. of Physics   Lafayette College\, Easton PA 18042

p5pRT commented 24 years ago

From @pudge

At 2.59 -0500 2000.04.03\, David Grove wrote​:

left it off and went on to the next configuration item. Even if\, get that\, even if "perl Makefile.PL POLLUTE=1" solves the problem\, 5.6.0 is still out of sync with CPAN.

That I don't see your name at the "current testers" page as tester of platforms Windows like at http​://testers.cpan.org/ surely must be a mistake\, right?

No. I wasn't aware of it.

Wow. It has been well-publicized and linked-to for quite some time now. Or so I thought. :/

p5pRT commented 24 years ago

From [Unknown Contact. See original ticket]

Andy Dougherty wrote​:

So what to do? Since you have apparently ignored this information the first time I presented it\, I will not waste my time repeating it\, but we built in both binary compatibility by default (so many folks don't have to rebuild anything) and source compatibility by two different optional routes. We did this about 16 months ago so CPAN authors might have time to catch up and adjust.

  I really have no argument with any of this\, but isn't it true   that something changed at almost the last minute that forced   the issue? My module compiled without changes as late as .670\,   possibly with RC1\, but required na => PL_na attention by RC2.

  -Norton

p5pRT commented 24 years ago

From @gsar

On Mon\, 03 Apr 2000 11​:18​:43 EDT\, Norton Allen wrote​:

Andy Dougherty wrote​:

So what to do? Since you have apparently ignored this information the first time I presented it\, I will not waste my time repeating it\, but we built in both binary compatibility by default (so many folks don't have to rebuild anything) and source compatibility by two different optional routes. We did this about 16 months ago so CPAN authors might have time to catch up and adjust.

I really have no argument with any of this\, but isn't it true that something changed at almost the last minute that forced the issue?

Not true. There were no changes to the PERL_POLLUTE default or any of the pollution variables since as far back as October 1999.

        My module compiled without changes as late as \.670\,

possibly with RC1\, but required na => PL_na attention by RC2.

I don't see anything about this in the archives. Did you report it via perlbug?

Sarathy gsar@​ActiveState.com

p5pRT commented 24 years ago

From [Unknown Contact. See original ticket]

Gurusamy Sarathy wrote​:

On Mon\, 03 Apr 2000 11​:18​:43 EDT\, Norton Allen wrote​:

Andy Dougherty wrote​:

So what to do? Since you have apparently ignored this information the first time I presented it\, I will not waste my time repeating it\, but we built in both binary compatibility by default (so many folks don't have to rebuild anything) and source compatibility by two different optional routes. We did this about 16 months ago so CPAN authors might have time to catch up and adjust.

I really have no argument with any of this\, but isn't it true that something changed at almost the last minute that forced the issue?

Not true. There were no changes to the PERL_POLLUTE default or any of the pollution variables since as far back as October 1999.

        My module compiled without changes as late as \.670\,

possibly with RC1\, but required na => PL_na attention by RC2.

I don't see anything about this in the archives. Did you report it via perlbug?

Sarathy gsar@​ActiveState.com

p5pRT commented 24 years ago

From [Unknown Contact. See original ticket]

[Sorry about the previous reply w/o new content. Finger slip]

Gurusamy Sarathy wrote​:

I really have no argument with any of this\, but isn't it true that something changed at almost the last minute that forced the issue?

Not true. There were no changes to the PERL_POLLUTE default or any of the pollution variables since as far back as October 1999.

  OK\, it is probably my week memory confusing releases that I   compiled versus releases I compiled and installed with all the   modules.

        My module compiled without changes as late as \.670\,

possibly with RC1\, but required na => PL_na attention by RC2.

I don't see anything about this in the archives. Did you report it via perlbug?

  I assumed (no doubt correctly) that I was coming across   something that was a long time in coming that needed to be   tended to\, so no\, I didn't report it. I knew the general issue\,   but wasn't following too closely the implementation calendar.  
  -Norton