-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512
Hi Ulrik!
On 07/01/2016 13:48, Ulrik Sandborg-Petersen wrote:
> I have fixed the bugs in t3.c. It turned out to be two issues:
>
> 1) size_t is unsigned, whereas a lot of code depended on various types
> being signed.
I reverted a few of the changes that you made, because in some interfaces
size_t is right.
I tried to remove many of the warnings, but it was not always possible to
remove them in a clean way, so I left some of them in the code.
> 2) In t3.c's "viterbi" function, I (or someone else) had introduced a
> bug somewhere between version 1.8.6 and 2.0.0.pre01, in how the "b"
> array was created. This led to all kinds of memory problems, hence the
> issue you saw with long sentences.
I have seen the changes and I confirm the performance are as expected with my
dataset.
However you introduced a memory leak with the change. So I reverted the change
and fixed the pointer math, so that the old solution is now working as it shoul
With respect to other issues that I am facing now is a few segmentation faults
in met (tagging, not training). The first of them is due to stack overflow.
If you have any idea about how to solve this issue, it is welcome.
Bests,
Giulio.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=NAqa
-----END PGP SIGNATURE-----
------------------------------------------------------------------------------
_______________________________________________
acopost-devel mailing list
acopost-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/acopost-devel
