Hi Stefan,

On Fri, 6 May 2016 21:49:34 +0200 Stefan Potyra <[email protected]>
wrote:
> Hi,
> 
> as promised a short status update:
> 
> There have been quite some changes upstream-wise, but unfortunately
it still
> doesn't build with current gcc version found in unstable(see TLDR
about the
> reason).

> Amd I've been once again accounted to do the upstream release...
> 
> So with my upstream hat on: I'd very much like to stay FAUmachine in
the
> archive, however with my DM hat, I think it's best to remove the
binary
> packages for now to not block anything.
> 
> Are you OK with this plan and/or do you have good advice for me?
> 
> Cheers,
>   Stefan.
>    
> 
> TLDR:
> FAUmachine uses a forked version of qemu to build a just-in-time
compiler. The
> idea from Fabrice Bellard is brilliant: Code the emulation of
assembly
> instructions in C and let gcc produce binary code for it. Then use a
script to
> mange this binary code to be usable as input as basic blocks for the
> jut-in-time-compiler.
> 
> Unfortunately, at this script is highly dependent on the produced
assembly code
> (e.g. just one return instruction per function) and this usually
breaks with
> every new gcc version :(.
> 
> This results in the current build breakage, and while this has been
fixed for
> the gcc version in stable, breakage occurs with the current gcc
version in
> unstable.
> 
> 

Thanks for your explanation.
That story seems to be like a maintainance
hell, it defintily makes it not easier, especially as there is already
gcc 6 on the horizon...
Frankly, IMHO this is very close to a case of
FTBFSIASW..
Sorry to be direct, but such an model will only fly with
distributions if upstream is following closely the gcc development.

As short-term measure, you can also force an older gcc version -- 
gcc4.9 is still in the archives: One successful upload will buy you time as 
then the libpng depdendency will be gone. However, be prepared to be nagged by 
then next developers as a bird told me gcc-4.9 is to be removed too. But 
targeting experimental maybe would be an compromise?

I'd be ok with binary only removal, but there is no guarantee that it works 
out: The binaryless-source might be accidentially de-crufted by our 
ftp-masters...
I'll retitle the bug accordingly; (Note: You should tag this "moreinfo" if you 
want to make sure that removal is not executed)


--
tobi

Reply via email to