fflush(NULL) clobbers input streams

fflush(NULL) clobbers input streams

[Achim Gratz]

Perl configure for 5.22.1 on Cygwin 2.4.0 said this:

--8<---------------cut here---------------start------------->8---
Checking how to flush all pending stdio output...
Your fflush(NULL) works okay for output streams.
Let's see if it clobbers input pipes...
fflush(NULL) seems to behave okay with input streams.
--8<---------------cut here---------------end--------------->8---

I'm reasonably certain that the intervening Perl versions were also this
way, but I didn't keep the log files around unfortunately.

Now, using Cygwin 2.8.x to configure Perl 5.22.4 I got this instead:

--8<---------------cut here---------------start------------->8---
Checking how to flush all pending stdio output...
Your fflush(NULL) works okay for output streams.
Let's see if it clobbers input pipes...
Ouch, fflush(NULL) clobbers input pipes!  We will not use it.
Good, at least fflush(stdin) seems to behave okay when stdin is a pipe.
Sigh. Flushing explicitly all the stdio streams doesn't work.
--8<---------------cut here---------------end--------------->8---

It seems that fflush is only supposed to discard buffered input data on
seekable files, but not pipes (per POSIX.1-2008), so this looks a
regression to me.


Regards,
Achim.
-- 
+<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk Blofeld]>+

Waldorf MIDI Implementation & additional documentation:
http://Synth.Stromeko.net/Downloads.html#WaldorfDocs

Re: fflush(NULL) clobbers input streams

[Corinna Vinschen]

[-- Attachment #1: Type: text/plain, Size: 1525 bytes --]

On Jul 17 19:53, Achim Gratz wrote:
> 
> Perl configure for 5.22.1 on Cygwin 2.4.0 said this:
> 
> --8<---------------cut here---------------start------------->8---
> Checking how to flush all pending stdio output...
> Your fflush(NULL) works okay for output streams.
> Let's see if it clobbers input pipes...
> fflush(NULL) seems to behave okay with input streams.
> --8<---------------cut here---------------end--------------->8---
> 
> I'm reasonably certain that the intervening Perl versions were also this
> way, but I didn't keep the log files around unfortunately.
> 
> Now, using Cygwin 2.8.x to configure Perl 5.22.4 I got this instead:
> 
> --8<---------------cut here---------------start------------->8---
> Checking how to flush all pending stdio output...
> Your fflush(NULL) works okay for output streams.
> Let's see if it clobbers input pipes...
> Ouch, fflush(NULL) clobbers input pipes!  We will not use it.
> Good, at least fflush(stdin) seems to behave okay when stdin is a pipe.
> Sigh. Flushing explicitly all the stdio streams doesn't work.
> --8<---------------cut here---------------end--------------->8---
> 
> It seems that fflush is only supposed to discard buffered input data on
> seekable files, but not pipes (per POSIX.1-2008), so this looks a
> regression to me.

Did you try to bisect the issue?


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Maintainer                 cygwin AT cygwin DOT com
Red Hat

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: fflush(NULL) clobbers input streams

[Corinna Vinschen]

[-- Attachment #1: Type: text/plain, Size: 1936 bytes --]

On Jul 18 11:54, Corinna Vinschen wrote:
> On Jul 17 19:53, Achim Gratz wrote:
> > 
> > Perl configure for 5.22.1 on Cygwin 2.4.0 said this:
> > 
> > --8<---------------cut here---------------start------------->8---
> > Checking how to flush all pending stdio output...
> > Your fflush(NULL) works okay for output streams.
> > Let's see if it clobbers input pipes...
> > fflush(NULL) seems to behave okay with input streams.
> > --8<---------------cut here---------------end--------------->8---
> > 
> > I'm reasonably certain that the intervening Perl versions were also this
> > way, but I didn't keep the log files around unfortunately.
> > 
> > Now, using Cygwin 2.8.x to configure Perl 5.22.4 I got this instead:
> > 
> > --8<---------------cut here---------------start------------->8---
> > Checking how to flush all pending stdio output...
> > Your fflush(NULL) works okay for output streams.
> > Let's see if it clobbers input pipes...
> > Ouch, fflush(NULL) clobbers input pipes!  We will not use it.
> > Good, at least fflush(stdin) seems to behave okay when stdin is a pipe.
> > Sigh. Flushing explicitly all the stdio streams doesn't work.
> > --8<---------------cut here---------------end--------------->8---
> > 
> > It seems that fflush is only supposed to discard buffered input data on
> > seekable files, but not pipes (per POSIX.1-2008), so this looks a
> > regression to me.
> 
> Did you try to bisect the issue?

The log doesn't indicate any obvious candidate.  We're talking about a
timeframe of 2016-01-15 - 2017-03-30 *iff* this already occurs on 2.8.0.

At least knowing the exact Cygwin version introducing the issue would
be a great help zeroing in on the culprit.

Oh, and a plain C testcase, of course...


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Maintainer                 cygwin AT cygwin DOT com
Red Hat

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: fflush(NULL) clobbers input streams

[Achim Gratz]

Corinna Vinschen writes:
> Oh, and a plain C testcase, of course...

Call that fflush.c:

--8<---------------cut here---------------start------------->8---
#include <stdio.h>
int
main(int argc, char **argv)
{
    char buf[1024];
    int i;
    char *bp = buf;
    while (1) {
        while ((i = getc(stdin)) != -1
               && (*bp++ = i) != '\n'
               && bp < &buf[1024])
        /* DO NOTHING */ ;
        *bp = '\0';
        fprintf(stdout, "%s", buf);
        fflush(NULL);
        if (i == -1)
            return 0;
        bp = buf;
    }
}
--8<---------------cut here---------------end--------------->8---

Then run:

gcc fflush.c -o fflush
cat fflush.c | ./fflush 2>/dev/null > fflush.out
diff fflush.c fflush.out

If things are working, fflush.c and fflush.out should be identical.
Currently only the first line makes it into fflush.out on Cygwin.


Regards,
Achim.
-- 
+<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk Blofeld]>+

Factory and User Sound Singles for Waldorf Blofeld:
http://Synth.Stromeko.net/Downloads.html#WaldorfSounds

Re: fflush(NULL) clobbers input streams

[Ken Brown]

On 7/18/2017 3:21 PM, Achim Gratz wrote:
> Corinna Vinschen writes:
>> Oh, and a plain C testcase, of course...
> 
> Call that fflush.c:
> 
> --8<---------------cut here---------------start------------->8---
> #include <stdio.h>
> int
> main(int argc, char **argv)
> {
>      char buf[1024];
>      int i;
>      char *bp = buf;
>      while (1) {
>          while ((i = getc(stdin)) != -1
>                 && (*bp++ = i) != '\n'
>                 && bp < &buf[1024])
>          /* DO NOTHING */ ;
>          *bp = '\0';
>          fprintf(stdout, "%s", buf);
>          fflush(NULL);
>          if (i == -1)
>              return 0;
>          bp = buf;
>      }
> }
> --8<---------------cut here---------------end--------------->8---
> 
> Then run:
> 
> gcc fflush.c -o fflush
> cat fflush.c | ./fflush 2>/dev/null > fflush.out
> diff fflush.c fflush.out
> 
> If things are working, fflush.c and fflush.out should be identical.
> Currently only the first line makes it into fflush.out on Cygwin.

I've checked that the 2017-03-08 snapshot is good and the 2017-03-10 
snapshot is bad.  I'll bisect further when I get a chance, but maybe 
that's already enough to pin down the problem.

Ken

Re: fflush(NULL) clobbers input streams

[Ken Brown]

On 7/18/2017 6:41 PM, Ken Brown wrote:
> On 7/18/2017 3:21 PM, Achim Gratz wrote:
>> Corinna Vinschen writes:
>>> Oh, and a plain C testcase, of course...
>>
>> Call that fflush.c:
>>
>> --8<---------------cut here---------------start------------->8---
>> #include <stdio.h>
>> int
>> main(int argc, char **argv)
>> {
>>      char buf[1024];
>>      int i;
>>      char *bp = buf;
>>      while (1) {
>>          while ((i = getc(stdin)) != -1
>>                 && (*bp++ = i) != '\n'
>>                 && bp < &buf[1024])
>>          /* DO NOTHING */ ;
>>          *bp = '\0';
>>          fprintf(stdout, "%s", buf);
>>          fflush(NULL);
>>          if (i == -1)
>>              return 0;
>>          bp = buf;
>>      }
>> }
>> --8<---------------cut here---------------end--------------->8---
>>
>> Then run:
>>
>> gcc fflush.c -o fflush
>> cat fflush.c | ./fflush 2>/dev/null > fflush.out
>> diff fflush.c fflush.out
>>
>> If things are working, fflush.c and fflush.out should be identical.
>> Currently only the first line makes it into fflush.out on Cygwin.
> 
> I've checked that the 2017-03-08 snapshot is good and the 2017-03-10 
> snapshot is bad.  I'll bisect further when I get a chance

Done:

44b1746a41921533d27aca414a9188314cb725b6 is the first bad commit
commit 44b1746a41921533d27aca414a9188314cb725b6
Author: Corinna Vinschen <corinna@vinschen.de>
Date:   Fri Mar 10 20:21:09 2017 +0100

     errno: Stop using _impure_ptr->_errno completely

     We use errno AKA _REENT->_errno since the last century and only set
     _impure_ptr->_errno for backward compat.  Stop that.  Also, remove
     the last check for _impure_ptr->_errno in Cygwin code.

Ken

Re: fflush(NULL) clobbers input streams

[Corinna Vinschen]

[-- Attachment #1: Type: text/plain, Size: 1523 bytes --]

On Jul 18 22:16, Ken Brown wrote:
> On 7/18/2017 6:41 PM, Ken Brown wrote:
> > On 7/18/2017 3:21 PM, Achim Gratz wrote:
> > > Corinna Vinschen writes:
> > > > Oh, and a plain C testcase, of course...
> > > 
> > > Call that fflush.c:
> > > [...]
> > > Then run:
> > > 
> > > gcc fflush.c -o fflush
> > > cat fflush.c | ./fflush 2>/dev/null > fflush.out
> > > diff fflush.c fflush.out
> > > 
> > > If things are working, fflush.c and fflush.out should be identical.
> > > Currently only the first line makes it into fflush.out on Cygwin.
> > 
> > I've checked that the 2017-03-08 snapshot is good and the 2017-03-10
> > snapshot is bad.  I'll bisect further when I get a chance
> 
> Done:
> 
> 44b1746a41921533d27aca414a9188314cb725b6 is the first bad commit
> commit 44b1746a41921533d27aca414a9188314cb725b6
> Author: Corinna Vinschen <corinna@vinschen.de>
> Date:   Fri Mar 10 20:21:09 2017 +0100
> 
>     errno: Stop using _impure_ptr->_errno completely
> 
>     We use errno AKA _REENT->_errno since the last century and only set
>     _impure_ptr->_errno for backward compat.  Stop that.  Also, remove
>     the last check for _impure_ptr->_errno in Cygwin code.
> 
> Ken

Achim, thanks for the testcase.  Ken, thanks for bisecting.  You're
my hero!

I reverted the patch and created new developer snapshots.


Thanks,
Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Maintainer                 cygwin AT cygwin DOT com
Red Hat

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

[GOLDSTAR] Re: fflush(NULL) clobbers input streams

[Corinna Vinschen]

[-- Attachment #1: Type: text/plain, Size: 1687 bytes --]

On Jul 19 10:45, Corinna Vinschen wrote:
> On Jul 18 22:16, Ken Brown wrote:
> > On 7/18/2017 6:41 PM, Ken Brown wrote:
> > > On 7/18/2017 3:21 PM, Achim Gratz wrote:
> > > > Corinna Vinschen writes:
> > > > > Oh, and a plain C testcase, of course...
> > > > 
> > > > Call that fflush.c:
> > > > [...]
> > > > Then run:
> > > > 
> > > > gcc fflush.c -o fflush
> > > > cat fflush.c | ./fflush 2>/dev/null > fflush.out
> > > > diff fflush.c fflush.out
> > > > 
> > > > If things are working, fflush.c and fflush.out should be identical.
> > > > Currently only the first line makes it into fflush.out on Cygwin.
> > > 
> > > I've checked that the 2017-03-08 snapshot is good and the 2017-03-10
> > > snapshot is bad.  I'll bisect further when I get a chance
> > 
> > Done:
> > 
> > 44b1746a41921533d27aca414a9188314cb725b6 is the first bad commit
> > commit 44b1746a41921533d27aca414a9188314cb725b6
> > Author: Corinna Vinschen <corinna@vinschen.de>
> > Date:   Fri Mar 10 20:21:09 2017 +0100
> > 
> >     errno: Stop using _impure_ptr->_errno completely
> > 
> >     We use errno AKA _REENT->_errno since the last century and only set
> >     _impure_ptr->_errno for backward compat.  Stop that.  Also, remove
> >     the last check for _impure_ptr->_errno in Cygwin code.
> > 
> > Ken
> 
> Achim, thanks for the testcase.  Ken, thanks for bisecting.  You're
> my hero!
> 
> I reverted the patch and created new developer snapshots.

I forgot the gold star.  Thanks again, Ken!  


Corinna



-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Maintainer                 cygwin AT cygwin DOT com
Red Hat

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: [GOLDSTAR] Re: fflush(NULL) clobbers input streams

[Ken Brown]

On 7/19/2017 1:03 PM, Corinna Vinschen wrote:
> On Jul 19 10:45, Corinna Vinschen wrote:
>> On Jul 18 22:16, Ken Brown wrote:
>>> On 7/18/2017 6:41 PM, Ken Brown wrote:
>>>> On 7/18/2017 3:21 PM, Achim Gratz wrote:
>>>>> Corinna Vinschen writes:
>>>>>> Oh, and a plain C testcase, of course...
>>>>>
>>>>> Call that fflush.c:
>>>>> [...]
>>>>> Then run:
>>>>>
>>>>> gcc fflush.c -o fflush
>>>>> cat fflush.c | ./fflush 2>/dev/null > fflush.out
>>>>> diff fflush.c fflush.out
>>>>>
>>>>> If things are working, fflush.c and fflush.out should be identical.
>>>>> Currently only the first line makes it into fflush.out on Cygwin.
>>>>
>>>> I've checked that the 2017-03-08 snapshot is good and the 2017-03-10
>>>> snapshot is bad.  I'll bisect further when I get a chance
>>>
>>> Done:
>>>
>>> 44b1746a41921533d27aca414a9188314cb725b6 is the first bad commit
>>> commit 44b1746a41921533d27aca414a9188314cb725b6
>>> Author: Corinna Vinschen <corinna@vinschen.de>
>>> Date:   Fri Mar 10 20:21:09 2017 +0100
>>>
>>>      errno: Stop using _impure_ptr->_errno completely
>>>
>>>      We use errno AKA _REENT->_errno since the last century and only set
>>>      _impure_ptr->_errno for backward compat.  Stop that.  Also, remove
>>>      the last check for _impure_ptr->_errno in Cygwin code.
>>>
>>> Ken
>>
>> Achim, thanks for the testcase.  Ken, thanks for bisecting.  You're
>> my hero!
>>
>> I reverted the patch and created new developer snapshots.
> 
> I forgot the gold star.  Thanks again, Ken!

Glad I could do it.  And thanks for the hero comment.  You made my day.

Ken

Re: [GOLDSTAR] Re: fflush(NULL) clobbers input streams

[Andrew Schulman]

> > Achim, thanks for the testcase.  Ken, thanks for bisecting.  You're
> > my hero!
> > 
> > I reverted the patch and created new developer snapshots.
> 
> I forgot the gold star.  Thanks again, Ken!  

Awarded!

https://cygwin.com/goldstars/#AG
https://cygwin.com/goldstars/#KB

Re: fflush(NULL) clobbers input streams

[Achim Gratz]

Corinna Vinschen writes:
> Achim, thanks for the testcase.

I really only pulled it out of the configure script.
:-)


Regards,
Achim.
-- 
+<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk Blofeld]>+

Wavetables for the Terratec KOMPLEXER:
http://Synth.Stromeko.net/Downloads.html#KomplexerWaves

Re: fflush(NULL) clobbers input streams

[Achim Gratz]

Ken Brown writes:
>> I've checked that the 2017-03-08 snapshot is good and the 2017-03-10
>> snapshot is bad.  I'll bisect further when I get a chance
>
> Done:
>
> 44b1746a41921533d27aca414a9188314cb725b6 is the first bad commit
> commit 44b1746a41921533d27aca414a9188314cb725b6
> Author: Corinna Vinschen <corinna-fFDxGeQsSAIb1SvskN2V4Q@public.gmane.org>
> Date:   Fri Mar 10 20:21:09 2017 +0100
>
>     errno: Stop using _impure_ptr->_errno completely
>
>     We use errno AKA _REENT->_errno since the last century and only set
>     _impure_ptr->_errno for backward compat.  Stop that.  Also, remove
>     the last check for _impure_ptr->_errno in Cygwin code.

Thank you!  Well, that would have been the last commit I'd suspected.

Andrew, are you perhaps having a goldstar in the vault for Mr. Brown?


Regards,
Achim.
-- 
+<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk Blofeld]>+

SD adaptations for Waldorf Q V3.00R3 and Q+ V3.54R2:
http://Synth.Stromeko.net/Downloads.html#WaldorfSDada

Re: fflush(NULL) clobbers input streams

[Achim Gratz]

Corinna Vinschen writes:
> On Jul 17 19:53, Achim Gratz wrote:
>> It seems that fflush is only supposed to discard buffered input data on
>> seekable files, but not pipes (per POSIX.1-2008), so this looks a
>> regression to me.
>
> Did you try to bisect the issue?

I've had to bin the build environment at work due to the disk running
out of space and the one at home went poof when the SSD it was on
croaked.  I haven't had the time yet to get it up again on the
replacement.

Anyway, if my hunch that 5.22.3 was not running into this issue is
correct, then 2.7.0 was the last released Cygwin version this test was
passing for.  I might try to get that running on of my home machines
again, but I can't really spend all that much time on it for the
foreseeable future unfortunately.


Regards,
Achim.
-- 
+<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk Blofeld]>+

SD adaptation for Waldorf microQ V2.22R2:
http://Synth.Stromeko.net/Downloads.html#WaldorfSDada