Bug#966525: dmagnetic FTBFS on i386: check-sixel failed

[dettus]

Hello everybody.


After Adrian was kind enough to give me a solution to my problem, I was 
able to apply it.


So my wonderful project dMagnetic can now be found with the release 
number 0.25-2 in

your mentors.debian.net system.

https://mentors.debian.net/package/dmagnetic/

Now, I would kindly ask you to upload it to the unstable servers.



Thomas

On 8/2/20 4:09 PM, Adrian Bunk wrote:

On Fri, Jul 31, 2020 at 06:02:03PM +0200, Thomas Dettbarn wrote:

Hello Adrian.

So, I patched my package, and uploaded to mentors.

I got a message with the subject "dmagnetic_0.25-1:
ACCEPTED into unstable."

It can be found here,

https://mentors.debian.net/package/dmagnetic/


but now I am a little bit concerned about the developer information
page: https://tracker.debian.org/pkg/dmagnetic it is still complaining
about i386 not working. Could you have a look?

Whatever you do on mentors does not have any effect in Debian,
unless it gets uploaded into the archive.

Version 0.25-1 of dmagnetic has been uploaded (by Adam Borowski).
This version cannot ever be changed in Debian.

You have to make a new revision 0.25-2 with the fix applied,
a changelog describing all changes like
   * Build with -ffloat-store on i386 to fix FTBFS. (Closes: #966525)
and that has to be uploaded to Debian for you.

I can sponsor such an upload if you upload it to mentors.


Thomas

cu
Adrian

Bug#966525: dmagnetic FTBFS on i386: check-sixel failed

[Adrian Bunk]

On Fri, Jul 31, 2020 at 06:02:03PM +0200, Thomas Dettbarn wrote:
> Hello Adrian.
> 
> So, I patched my package, and uploaded to mentors.
> 
> I got a message with the subject "dmagnetic_0.25-1:
> ACCEPTED into unstable."
> 
> It can be found here,
> 
> https://mentors.debian.net/package/dmagnetic/
> 
> 
> but now I am a little bit concerned about the developer information
> page: https://tracker.debian.org/pkg/dmagnetic it is still complaining
> about i386 not working. Could you have a look?

Whatever you do on mentors does not have any effect in Debian,
unless it gets uploaded into the archive.

Version 0.25-1 of dmagnetic has been uploaded (by Adam Borowski).
This version cannot ever be changed in Debian.

You have to make a new revision 0.25-2 with the fix applied,
a changelog describing all changes like
  * Build with -ffloat-store on i386 to fix FTBFS. (Closes: #966525)
and that has to be uploaded to Debian for you.

I can sponsor such an upload if you upload it to mentors.

> Thomas

cu
Adrian

Bug#966525: dmagnetic FTBFS on i386: check-sixel failed

[Thomas Dettbarn]

Hello Adrian.

So, I patched my package, and uploaded to mentors.

I got a message with the subject "dmagnetic_0.25-1:
ACCEPTED into unstable."

It can be found here,

https://mentors.debian.net/package/dmagnetic/


but now I am a little bit concerned about the developer information
page: https://tracker.debian.org/pkg/dmagnetic it is still complaining
about i386 not working. Could you have a look?



Thomas


On 7/30/20 4:29 PM, Adrian Bunk wrote:

On Thu, Jul 30, 2020 at 03:47:49PM +0200, Thomas Dettbarn wrote:

...
One last thing... Why did your patch say "ifneq... i386"?

ifneq (,$(filter $(DEB_HOST_ARCH), i386))
^
This compares the empty string with the filter result.

ifeq ($(DEB_HOST_ARCH), i386)
would to the same in this case.

I am used to the filter way of writing it since it allows listing
several architectures like
ifneq (,$(filter $(DEB_HOST_ARCH), i386 mipsel hppa))


Thomas

cu
Adrian

Bug#966525: dmagnetic FTBFS on i386: check-sixel failed

[Thomas Dettbarn]

 
 
  
   https://mentors.debian.net/package/dmagnetic/
   
  
    
   
  
   Okay. I applied your patch. Please have a look.
   
   
   
Adrian Bunk  hat am 30.07.2020 16:29 geschrieben:

   
 

   
 

   
On Thu, Jul 30, 2020 at 03:47:49PM +0200, Thomas Dettbarn wrote:

   
>...



 One last thing... Why did your patch say "ifneq... i386"?
 

   
ifneq (,$(filter $(DEB_HOST_ARCH), i386))

   
^

   
This compares the empty string with the filter result.

   
 

   
ifeq ($(DEB_HOST_ARCH), i386)

   
would to the same in this case.

   
 

   
I am used to the filter way of writing it since it allows listing

   
several architectures like

   
ifneq (,$(filter $(DEB_HOST_ARCH), i386 mipsel hppa))

   
 



 Thomas
 

   
cu

   
Adrian

  
 

Bug#966525: dmagnetic FTBFS on i386: check-sixel failed

[Adrian Bunk]

On Thu, Jul 30, 2020 at 03:47:49PM +0200, Thomas Dettbarn wrote:
>...
> One last thing... Why did your patch say "ifneq... i386"?

ifneq (,$(filter $(DEB_HOST_ARCH), i386))
   ^
This compares the empty string with the filter result.

ifeq ($(DEB_HOST_ARCH), i386)
would to the same in this case.

I am used to the filter way of writing it since it allows listing 
several architectures like
ifneq (,$(filter $(DEB_HOST_ARCH), i386 mipsel hppa))

> Thomas

cu
Adrian

Bug#966525: dmagnetic FTBFS on i386: check-sixel failed

[Thomas Dettbarn]

Man, I hate floats... I try to avoid them as much as I can,
because I am dreading the butterfly effect. And, this time around
I was lazy, and one of the butterflies kicked me.

Funny thing about the m68k: My project ACTUALLY has an m68k virtual machine 
inside. ;)


Okay, so: 

This release (0.25) will be fixed in the debian/rules file.
Next release (0.26) will be fixed in the software.

Thank you for pointing it out!


One last thing... Why did your patch say "ifneq... i386"?



Thomas
> Adrian Bunk  hat am 30.07.2020 15:15 geschrieben:
> 
>  
> On Thu, Jul 30, 2020 at 02:46:55PM +0200, Thomas Dettbarn wrote:
> > Hello Adrian!
> 
> Hi Thomas!
> 
> >...
> > Apparently, it only happens on 
> > i386, but I fear that it might occur on other architectures as
> > well.
> 
> For being able to run on old hardware, only the old x87 FPU is used on
> the Debian i386 port.
> 
> The x87 FPU internally uses 80bit precision instead of 64bit precision.
> The results are "better", but different from IEEE floating point.
> 
> The only other Debian architecture that would have such a problem
> would be m68k, but builds there are with nocheck.
> 
> > Thomas
> 
> cu
> Adrian

Bug#966525: dmagnetic FTBFS on i386: check-sixel failed

[Adrian Bunk]

On Thu, Jul 30, 2020 at 02:46:55PM +0200, Thomas Dettbarn wrote:
> Hello Adrian!

Hi Thomas!

>...
> Apparently, it only happens on 
> i386, but I fear that it might occur on other architectures as
> well.

For being able to run on old hardware, only the old x87 FPU is used on
the Debian i386 port.

The x87 FPU internally uses 80bit precision instead of 64bit precision.
The results are "better", but different from IEEE floating point.

The only other Debian architecture that would have such a problem
would be m68k, but builds there are with nocheck.

> Thomas

cu
Adrian

Bug#966525: dmagnetic FTBFS on i386: check-sixel failed

[Thomas Dettbarn]

Hello Adrian!


Hmm... I wonder what is going on there?
May I bother you to send me the output of


echo "Dreye"| ./dMagnetic -ini dMagnetic.ini \
-vmode "sixel" -vcols 300 -vrows 300 -vecho \
-sres 1024x768 -mag testcode/minitest.mag >fails.txt



running on an i386 computer with Debian? That would help me to
understand what the issue is. Apparently, it only happens on 
i386, but I fear that it might occur on other architectures as
well.

Thomas

> Adrian Bunk  hat am 30.07.2020 08:04 geschrieben:
> 
>  
> Source: dmagnetic
> Version: 0.25-1
> Severity: serious
> Tags: ftbfs patch
> 
> https://buildd.debian.org/status/fetch.php?pkg=dmagnetic&arch=i386&ver=0.25-1&stamp=1595804868&raw=0
> 
> ...
> Initializing the 68K version 0 VM
> check-high_ansi OK
> *** dMagnetic 0.25
> *** Use at your own risk
> *** (C)opyright 2020 by det...@dettus.net
> *
> 
> Using .ini file: dMagnetic.ini
> Warning! -mag given, but not -gfx. Deducing filename
> Initializing the 68K version 0 VM
> check-sixel failed
> make[1]: *** [checks.mk:104: check-sixel] Error 1
> 
> 
> If exact floating points are required, the following fixes it:
> 
> --- debian/rules.old  2020-07-30 09:01:53.400678013 +0300
> +++ debian/rules  2020-07-30 09:02:37.400712152 +0300
> @@ -3,6 +3,9 @@
>  export DH_VERBOSE = 1
>  export DEB_BUILD_MAINT_OPTIONS = hardening=+all
>  export DEB_CFLAGS_MAINT_APPEND  = -Wall -pedantic
> +ifneq (,$(filter $(DEB_HOST_ARCH), i386))
> +  export DEB_CFLAGS_MAINT_APPEND += -ffloat-store
> +endif
>  export SHA256_CMD=sha256sum
>  #export DEB_LDFLAGS_MAINT_APPEND = -Wl,--as-needed
>  #export DEB_RULES_REQUIRES_ROOT = ""

Bug#966525: dmagnetic FTBFS on i386: check-sixel failed

[Thomas Dettbarn]

Edit:

Okay, now I see the problem: It is the only place in the code
where I used floats. Hmmm. Bummer... You said that your patch
helps with that? Then I will add it to the debian package. 

For future releases, I will get rid of the floating point code.
It is ugly anyways.



Thomas


> Thomas Dettbarn  hat am 30.07.2020 14:46 geschrieben:
> 
>  
> Hello Adrian!
> 
> 
> Hmm... I wonder what is going on there?
> May I bother you to send me the output of
> 
> 
> echo "Dreye"| ./dMagnetic -ini dMagnetic.ini \
> -vmode "sixel" -vcols 300 -vrows 300 -vecho \
> -sres 1024x768 -mag testcode/minitest.mag >fails.txt
> 
> 
> 
> running on an i386 computer with Debian? That would help me to
> understand what the issue is. Apparently, it only happens on 
> i386, but I fear that it might occur on other architectures as
> well.
> 
> Thomas
> 
> > Adrian Bunk  hat am 30.07.2020 08:04 geschrieben:
> > 
> >  
> > Source: dmagnetic
> > Version: 0.25-1
> > Severity: serious
> > Tags: ftbfs patch
> > 
> > https://buildd.debian.org/status/fetch.php?pkg=dmagnetic&arch=i386&ver=0.25-1&stamp=1595804868&raw=0
> > 
> > ...
> > Initializing the 68K version 0 VM
> > check-high_ansi OK
> > *** dMagnetic 0.25
> > *** Use at your own risk
> > *** (C)opyright 2020 by det...@dettus.net
> > *
> > 
> > Using .ini file: dMagnetic.ini
> > Warning! -mag given, but not -gfx. Deducing filename
> > Initializing the 68K version 0 VM
> > check-sixel failed
> > make[1]: *** [checks.mk:104: check-sixel] Error 1
> > 
> > 
> > If exact floating points are required, the following fixes it:
> > 
> > --- debian/rules.old2020-07-30 09:01:53.400678013 +0300
> > +++ debian/rules2020-07-30 09:02:37.400712152 +0300
> > @@ -3,6 +3,9 @@
> >  export DH_VERBOSE = 1
> >  export DEB_BUILD_MAINT_OPTIONS = hardening=+all
> >  export DEB_CFLAGS_MAINT_APPEND  = -Wall -pedantic
> > +ifneq (,$(filter $(DEB_HOST_ARCH), i386))
> > +  export DEB_CFLAGS_MAINT_APPEND += -ffloat-store
> > +endif
> >  export SHA256_CMD=sha256sum
> >  #export DEB_LDFLAGS_MAINT_APPEND = -Wl,--as-needed
> >  #export DEB_RULES_REQUIRES_ROOT = ""

Bug#966525: dmagnetic FTBFS on i386: check-sixel failed

[Adrian Bunk]

Source: dmagnetic
Version: 0.25-1
Severity: serious
Tags: ftbfs patch

https://buildd.debian.org/status/fetch.php?pkg=dmagnetic&arch=i386&ver=0.25-1&stamp=1595804868&raw=0

...
Initializing the 68K version 0 VM
check-high_ansi OK
*** dMagnetic 0.25
*** Use at your own risk
*** (C)opyright 2020 by det...@dettus.net
*

Using .ini file: dMagnetic.ini
Warning! -mag given, but not -gfx. Deducing filename
Initializing the 68K version 0 VM
check-sixel failed
make[1]: *** [checks.mk:104: check-sixel] Error 1


If exact floating points are required, the following fixes it:

--- debian/rules.old2020-07-30 09:01:53.400678013 +0300
+++ debian/rules2020-07-30 09:02:37.400712152 +0300
@@ -3,6 +3,9 @@
 export DH_VERBOSE = 1
 export DEB_BUILD_MAINT_OPTIONS = hardening=+all
 export DEB_CFLAGS_MAINT_APPEND  = -Wall -pedantic
+ifneq (,$(filter $(DEB_HOST_ARCH), i386))
+  export DEB_CFLAGS_MAINT_APPEND += -ffloat-store
+endif
 export SHA256_CMD=sha256sum
 #export DEB_LDFLAGS_MAINT_APPEND = -Wl,--as-needed
 #export DEB_RULES_REQUIRES_ROOT = ""