Re: Bug in cygwin with select and master pty

Re: Bug in cygwin with select and master pty

[cgf]

In article <001e01c049b7$7f159ea0$21c9ca95@mow.siemens.ru>,
Andrej Borsenkow <Andrej.Borsenkow@mow.siemens.ru> wrote:
>> } About select - so far there seems to be no known problems under cygwin
>> } (that does not mean, there are no problems).
>>
>>OK, if there are no known problems with select, then the test on line
>>1376 can change from `ret <= 0' to `ret < 0' -- and in that case it's
>>possible that we could split read_poll() into two cases, HAVE_SELECT v.
>>all the other code that's there.  I won't do that for now, though.
>
>Life is fun.  5 minutes later I discovered a bug that made blocking
>multiline reads (with or without pattern) hang at the end of the first
>line.  I sent sort of a patch to cygwin list, but I am not sure if it
>really does the right thing (it fixed the bug I've seen, O.K.  The
>question is whether it introduced another bugs :-)

I didn't see any followup to this but the new (1.1.5) release of cygwin
should not have this problem.

>Anyway, fiddling with currently released versions of cygwin is pretty
>useless wrt to zpty (all of them are buggy in some way).  Unfortunately
>(to us) zsh seems to be the only program that has these problems.  With
>all implications.

Have all of these problems been reported to the cygwin mailing list?  Are
there still problems in 1.1.5?

>The pgrp problem still remains.

What is the pgrp problem?  Was this reported to the cygwin mailing list?

cgf

RE: Bug in cygwin with select and master pty

[Andrej Borsenkow]

> -----Original Message-----
> From: cgf@redhat.com [mailto:cgf@redhat.com]
> Sent: Sunday, November 12, 2000 11:20 PM
> To: zsh-workers@sunsite.auc.dk
> Subject: Re: Bug in cygwin with select and master pty
>
>
> In article <001e01c049b7$7f159ea0$21c9ca95@mow.siemens.ru>,
> Andrej Borsenkow <Andrej.Borsenkow@mow.siemens.ru> wrote:
> >> } About select - so far there seems to be no known problems under cygwin
> >> } (that does not mean, there are no problems).
> >>
> >>OK, if there are no known problems with select, then the test on line
> >>1376 can change from `ret <= 0' to `ret < 0' -- and in that case it's
> >>possible that we could split read_poll() into two cases, HAVE_SELECT v.
> >>all the other code that's there.  I won't do that for now, though.
> >
> >Life is fun.  5 minutes later I discovered a bug that made blocking
> >multiline reads (with or without pattern) hang at the end of the first
> >line.  I sent sort of a patch to cygwin list, but I am not sure if it
> >really does the right thing (it fixed the bug I've seen, O.K.  The
> >question is whether it introduced another bugs :-)
>
> I didn't see any followup to this but the new (1.1.5) release of cygwin
> should not have this problem.
>

Chris, are you hearing on this list too?

> >Anyway, fiddling with currently released versions of cygwin is pretty
> >useless wrt to zpty (all of them are buggy in some way).  Unfortunately
> >(to us) zsh seems to be the only program that has these problems.  With
> >all implications.
>
> Have all of these problems been reported to the cygwin mailing list?  Are
> there still problems in 1.1.5?
>

Yes. Build problem with termcap, EOF problem with pty, one more problem in
pre-1.1.5 snapshots and the above select problem.

The select problem is not fixed in 1.1.5-6.

> >The pgrp problem still remains.
>
> What is the pgrp problem?  Was this reported to the cygwin mailing list?
>

Some zpty code was changed and now zsh cannot attach to allocated pty (with
error message cannot set pgrp). No, this was not reported to cygwin list
because

- the problem resulted from zsh code change, so I first had to make sure it
was cygwin problem
- I have had enough "this is open source" replies. I'm sorry to say, but
reporting anything to cygwin list is pretty useless unless one can either
provide a patch or two-line program to reproduce the problem. I currently
cannot do either.

If you do not mind to update zsh CVS, you can see the last problem. Just do
something like

zmodload zsh/zpty
zpty sh /bin/sh

and now we have:

mw1g017@MW1G17C% zmodload zsh/zpty
mw1g017@MW1G17C% zpty sh /bin/sh
mw1g017@MW1G17C% ps -l
      PID    PPID    PGID     WINPID TTY  UID    STIME COMMAND
     1592       1    1592       1592   0 1006 14:05:53 /usr/bin/zsh
     1596    1592    1596       1596  -1 1006 14:06:14 /usr/bin/zsh
     1620    1592    1620       1624   0 1006 14:06:33 /usr/bin/ps

where provess 1596 is a child that tried to attach to allocated pty. If you
read the output you get:

mw1g017@MW1G17C% zpty -r sh foo \*$'\n'
mw1g017@MW1G17C% print -r $foo
zsh: can't set tty pgrp: not owner

This did work before last change; the main change was that zsh now opens both
master and slave fd's of pty with O_NOCTTY. This works on Unix.

-andrej

Re: Bug in cygwin with select and master pty

[Chris Faylor]

On Mon, Nov 13, 2000 at 02:12:50PM +0300, Andrej Borsenkow wrote:
>> I didn't see any followup to this but the new (1.1.5) release of cygwin
>> should not have this problem.
>
>Chris, are you hearing on this list too?

I've been reading the zsh-workers list for about four or five years, yes.

>> >Anyway, fiddling with currently released versions of cygwin is pretty
>> >useless wrt to zpty (all of them are buggy in some way).  Unfortunately
>> >(to us) zsh seems to be the only program that has these problems.  With
>> >all implications.
>>
>> Have all of these problems been reported to the cygwin mailing list?  Are
>> there still problems in 1.1.5?
>
>Yes. Build problem with termcap, EOF problem with pty, one more problem in
>pre-1.1.5 snapshots and the above select problem.

None of these statements are adequate to track any problems down, unfortunately.

>The select problem is not fixed in 1.1.5-6.

1.1.5-6 is history.

cgf

RE: Bug in cygwin with select and master pty

[Andrej Borsenkow]

> >
> >Yes. Build problem with termcap, EOF problem with pty, one more problem in
> >pre-1.1.5 snapshots and the above select problem.
> 
> None of these statements are adequate to track any problems down, 
> unfortunately.
> 

All of them are fixed currently. 

-andrej

Bug in cygwin with select and master pty

[Andrej Borsenkow]

>
> } About select - so far there seems to be no known problems under cygwin
> } (that does not mean, there are no problems).
>
> OK, if there are no known problems with select, then the test on line
> 1376 can change from `ret <= 0' to `ret < 0' -- and in that case it's
> possible that we could split read_poll() into two cases, HAVE_SELECT v.
> all the other code that's there.  I won't do that for now, though.
>

Life is fun. 5 minutes later I discovered a bug that made blocking multiline
reads (with or without pattern) hang at the end of the first line. I sent sort
of a patch to cygwin list, but I am not sure if it really does the right thing
(it fixed the bug I've seen, O.K. The question is whether it introduced
another bugs :-)

Anyway, fiddling with currently released versions of cygwin is pretty useless
wrt to zpty (all of them are buggy in some way). Unfortunately (to us) zsh
seems to be the only program that has these problems. With all implications.

The pgrp problem still remains.

-andrej