Hi Hector,

you're reporting three problems.

Problem #1:

>   While trying to solve #598370, which seems to be related to tthsum.
> I tried 1.2.0 version you posted, but I get a failure on little endian
> test on `armel' (ARM EABI 32 bit Architecture)

tthsum-1.0 was once upon a time based on the code from DC++ and was
therefore written in C++. tthsum-1.1 was mostly rewritten from scratch,
using the reference tiger.c implementation. It should bear little
resemblance to the DC++ original (as used in valknut/dclib).

I find it odd that the tiger_test and thex_test found in tthsum-1.2 would
return OK while the tthsum on dclib-0.3.23.tar.bz2 returns something other
than BRNGRD2NBXBLCJHRE27KNGK4CDFTTB65MNZIXHI. Did you really use
tthsum-1.2+ there? (Ok, the thex_test is limited to four very short bits
of data, so it's possible your problem is not caught.)


Problem #2:

> (sid)zu...@abel:~/tthsum-1.2.0$ make
> gcc -DNDEBUG -D_FILE_OFFSET_BITS=64 -DUSE_TEXTS  -fomit-frame-pointer
> -O3  -Wall  -c tthsum.c -o obj-unix/tthsum.o
...
>>>> endian_test.c:90: Little endian check fails on longs
> + endian_test: ...F.
...
> + tiger_test: ...
> + thex_test: .

I would not expect a long to not be 32 bits on your ARM. Either that, or
the (char) is unsigned and the test fails for that reason.

In any case. The test is broken. Assuming that a long is 32-bits is wrong.
Removing the bad test fixes this problem :)


Problem #3:

> I have also seen 1.2.1 release on site, which gives me (also on armel):
>
> (sid)zu...@abel:~/tthsum-1.2.1$ make
> gcc -DNDEBUG -D_FILE_OFFSET_BITS=64 -DUSE_TEXTS  -march=native -O3
> -Wall  -c tthsum.c -o obj-unix/tthsum.o
> tthsum.c:1: error: bad value (native) for -march= switch
> make: *** [obj-unix/tthsum.o] Error 1
>
> Tweaking Makefile for -DNDEBUG -D_FILE_OFFSET_BITS=32 and also,
> replacing "-march=native -O3"  by "-O2" fails on the same test.

Okay. This just means that the -march= native flag is not supported on
your gcc. Removing it from the Makefile is all there is to it.


Followup:

If you were using tthsum 1.2+ when you got the tthsum mismatch on
dclib-0.3.23.tar.bz2, please try
http://tthsum.devs.nu/pkg/hector517619.patch and report whether it gives
errors on the thex_test.


Regards,
Walter Doekes



--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org

Reply via email to