Re: [ANNOUNCEMENT] llvm 3.7.1-2

Re: [ANNOUNCEMENT] llvm 3.7.1-2

[lloyd.wood]

Thanks, picking up 3.7.1-2 from the mirrors.kernel.org site (as it's not mirrored everywhere yet) resolved the 32-bit OpenGL crashing issues with SaVi and Geomview that I reported in May; texturemapping is back.

(I've no idea how 64-bit OpenGL is doing, as unique Cygwin piping issues there prevent SaVi and Geomview from communicating.)

regards

Lloyd Wood
http://savi.sf.net


***
The following packages have been uploaded to the Cygwin distribution: * llvm-3.7.1-2
* llvm-doc-3.7.1-2
* libllvm3.7-3.7.1-2
* libllvm-devel-3.7.1-2 The LLVM Core libraries provide a modern source- and target-independent optimizer, along with code generation support for many popular CPUs (as well as some less common ones!) These libraries are built around a well specified code representation known as the LLVM intermediate representation. This release includes a backport and fix for RuntimeDyldCOFFI386 from Jon Turney in order to fix OpenGL software rendering on 32-bit Cygwin.

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: [ANNOUNCEMENT] llvm 3.7.1-2

[Yaakov Selkowitz]

On 6/9/2016 8:18 PM, lloyd.wood@yahoo.co.uk wrote:
> Thanks, picking up 3.7.1-2 from the mirrors.kernel.org site (as it's
> not mirrored everywhere yet) resolved the 32-bit OpenGL crashing issues
> with SaVi and Geomview that I reported in May; texturemapping is back.

Glad to hear it.

> (I've no idea how 64-bit OpenGL is doing, as unique Cygwin piping issues
> there prevent SaVi and Geomview from communicating.)

64-bit was fine all along; that's what I actually use, so I would have 
noticed earlier if it hadn't been working.

-- 
Yaakov

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: [ANNOUNCEMENT] llvm 3.7.1-2

[Jon Turney]

On 10/06/2016 02:18, lloyd.wood@yahoo.co.uk wrote:
> Thanks, picking up 3.7.1-2 from the mirrors.kernel.org site (as it's
> not mirrored everywhere yet) resolved the 32-bit OpenGL crashing
> issues with SaVi and Geomview that I reported in May; texturemapping
> is back.

You're welcome.

> (I've no idea how 64-bit OpenGL is doing, as unique Cygwin piping
> issues there prevent SaVi and Geomview from communicating.)

I guess this is the same 'piping problem' I asked for more details on 
back in [1]?

Can you please provide more information?

[1] https://cygwin.com/ml/cygwin/2013-07/msg00675.html

-- 
Jon Turney
Volunteer Cygwin/X X Server maintainer

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Piping on 64-bit Cygwin was Re: [ANNOUNCEMENT] llvm 3.7.1-2

[lloyd.wood]

Yes, it's the same piping problem of three years ago.

At the time the Geomview community wondered if this was
the shape of things to come for piping and IPC in general,
and if Geomview would need fixing for all platforms.
But since 32-bit and 64-bit Cygwin differed, and OpenGL
didn't work on 64-bit Cygwin either, making 32-bit the
obvious choice to use, masterful inaction was the
result. Wait and see how 64-bit Cygwin shakes out...


Three years on, 64-bit Cygwin is still the outlier;
piping works everywhere else, 32- and 64-bit. It would
be good to have 64-bit Cygwin work with Geomview and
its modules, particularly since 32-bit OpenGL seems less
than reliable (as also shown by the weird things seen
on Raspbian on the Pi).


It would also be good if Geomview was in the Cygwin
installer, as compiling it is not easy for many; it's
already a Debian package.


But I don't personally have the technical piping or
Cygwin-specific knowledge to troubleshoot this one,
other than providing and maintaining the installation
hints at


http://sat-net.com/L.Wood/software/SaVi/building-under-Windows/

thanks

Lloyd Wood
http://savi.sf.net/
(that screenshot comes from 32-bit Cygwin, btw.) 


----- Original Message -----
From: Jon Turney <jon.turney@dronecode.org.uk>
To: cygwin@cygwin.com
Cc: lloyd.wood@yahoo.co.uk
Sent: Monday, 20 June 2016, 22:58
Subject: Re: [ANNOUNCEMENT] llvm 3.7.1-2

On 10/06/2016 02:18, lloyd.wood@yahoo.co.uk wrote:
> Thanks, picking up 3.7.1-2 from the mirrors.kernel.org site (as it's
> not mirrored everywhere yet) resolved the 32-bit OpenGL crashing
> issues with SaVi and Geomview that I reported in May; texturemapping
> is back.

You're welcome.

> (I've no idea how 64-bit OpenGL is doing, as unique Cygwin piping
> issues there prevent SaVi and Geomview from communicating.)

I guess this is the same 'piping problem' I asked for more details on 
back in [1]?

Can you please provide more information?

[1] https://cygwin.com/ml/cygwin/2013-07/msg00675.html


-- 
Jon Turney
Volunteer Cygwin/X X Server maintainer

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Piping on 64-bit Cygwin was Re: [ANNOUNCEMENT] llvm 3.7.1-2

[Warren Young]

On Jun 20, 2016, at 10:53 PM, lloyd.wood@yahoo.co.uk wrote:
> 
> Yes, it's the same piping problem of three years ago.

…where you were asked to provide a simple test case for the problem, instead of “compile admittedly difficult-to-build package Geomview and use it against one of the most complicated packages in Cygwin, OpenGL.”

Here’s what a simple test case looks like:

  $ dd if=/dev/urandom bs=4k count=4m | 
    gpg -c --force-mdc | 
    gpg -d > /dev/null

Type the same passphrase three times: twice to verify it for encryption, and once for decryption.  The fact that this runs without errors proves that Cygwin 64 is capable of reliably transporting 16 GiB of data through a pipe without corruption.  

Try it on your system.  If it gives an error, it shows that the problem is system-specific, and thus possibly BLODA or some other platform issue.

If not, then please explain how your problem differs from the demonstration here.

Have you ruled out Cygwin’s OpenGL package, such as by running it over X11 to a remote system running a different implementation?
--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Piping on 64-bit Cygwin, STC found

[Warren Young]

On Jun 21, 2016, at 3:57 PM, Warren Young <wyml@etr-usa.com> wrote:
> 
> Here’s what a simple test case looks like:
> 
>  $ dd if=/dev/urandom bs=4k count=4m | 
>    gpg -c --force-mdc | 
>    gpg -d > /dev/null

I seem to have stumbled upon the actual STC.  Just increase those values, launch two MinTTY terminals, and try to use Cygwin in the second while the GPG pipeline is running.  All will be well for a while, but eventually the second window will start to behave strangely.

I used 64k and 64M (capitalization is important to dd), and the failure I observed is that “man” commands wouldn’t always return you to the shell properly after exiting from the piped-to less process.

Some kind of memory leak, perhaps?
--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Piping on 64-bit Cygwin, STC found

[Warren Young]

On Jun 21, 2016, at 4:40 PM, Warren Young <wyml@etr-usa.com> wrote:
> 
> On Jun 21, 2016, at 3:57 PM, Warren Young <wyml@etr-usa.com> wrote:
>> 
>> Here’s what a simple test case looks like:
>> 
>> $ dd if=/dev/urandom bs=4k count=4m | 
>>   gpg -c --force-mdc | 
>>   gpg -d > /dev/null
> 
> I seem to have stumbled upon the actual STC.

Here’s a screen shot of the error the second MinTTY gives on trying to exit:

  https://imgur.com/5RLi4ie

The “man bash” is the command I tried to give, and the “vi” command never executed, even after closing the first MinTTY window.  The defunct “less” process is of course a child of the “man” process.

> Just increase those values...I used 64k and 64M

To be clear, I mean bs=64k count=64M, for 4 TiB of transfer.  This ran for about half an hour before it started causing problems in the other MinTTY window here.  I didn’t write down the amount transferred at that time, unfortunately, though dd did report it.  Almost certainly many times more than available RAM at any rate.

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Piping on 64-bit Cygwin, STC found

[Andrey Repin]

Greetings, Warren Young!

> I used 64k and 64M (capitalization is important to dd),

You aren't confusing DD with some other tool?

$ dd if=/dev/urandom of=/dev/null bs=64K count=1
1+0 records in
1+0 records out
65536 bytes (66 kB, 64 KiB) copied, 0,0183668 s, 3,6 MB/s

$ dd if=/dev/urandom of=/dev/null bs=64k count=1
1+0 records in
1+0 records out
65536 bytes (66 kB, 64 KiB) copied, 0,00557431 s, 11,8 MB/s

$ dd if=/dev/urandom of=/dev/null bs=64kb count=1
dd: invalid number: ‘64kb’

$ dd if=/dev/urandom of=/dev/null bs=64kB count=1
1+0 records in
1+0 records out
64000 bytes (64 kB, 62 KiB) copied, 0,00562949 s, 11,4 MB/s

$ dd if=/dev/urandom of=/dev/null bs=64KB count=1
1+0 records in
1+0 records out
64000 bytes (64 kB, 62 KiB) copied, 0,00685924 s, 9,3 MB/s

$ dd --version
dd (coreutils) 8.25
Packaged by Cygwin (8.25-3)


-- 
With best regards,
Andrey Repin
Wednesday, June 22, 2016 02:08:29

Sorry for my terrible english...

Re: Piping on 64-bit Cygwin, STC found

[Warren Young]

On Jun 21, 2016, at 5:12 PM, Andrey Repin <anrdaemon@yandex.ru> wrote:
> 
> Greetings, Warren Young!
> 
>> I used 64k and 64M (capitalization is important to dd),
> 
> You aren't confusing DD with some other tool?

Try testing with “M”, not “k”.

Technically speaking, if it’s going to be picky about m vs M, it shouldn’t accept “K”, only “k”:

  https://en.wikipedia.org/wiki/Metric_prefix
--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: Piping on 64-bit Cygwin, STC found

[lloyd.wood]

> I seem to have stumbled upon the actual STC. 

That is not the STC for SaVi and Geomview under 64-bit Cygwin.

The problem for SaVi and Geomview is in INITIALISING any
piping, as previously described. They simply cannot communicate
at all under 64-bit Cygwin, while they can under 32-bit Cygwin.

You appear to have demonstrated that Cygwin piping fails
when used heavily. Unfortunately, that is not news, and
has always been so...

Lloyd Wood
http://savi.sf.net/

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple