From jordigh at octave.org  Thu May  3 16:05:57 2012
From: jordigh at octave.org (=?UTF-8?Q?Jordi_Guti=C3=A9rrez_Hermoso?=)
Date: Thu, 3 May 2012 16:05:57 -0400
Subject: [MINC-development] Problem with casting hid_t to pointer in
	hdf_convenience.c
Message-ID: <CAPHS2gznP-n40AUUcmZReT18wZpWtHoHp3DJ_r8c5PVOiz4hzQ@mail.gmail.com>

Hi, I'm trying to compile this MINC 2 git repo, at revision
2c967639bb:

    https://github.com/BIC-MNI/minc

However, I'm facing the following warnings, which I am pretty sure are
not safe to ignore:

    cc1: warnings being treated as errors

    minc/libsrc/hdf_convenience.c: In function ?hdf_attput?:
    minc/libsrc/hdf_convenience.c:1110: error: cast to pointer from
    integer of different size

    minc/libsrc/hdf_convenience.c: In function ?hdf_copy_attr?:
    minc/libsrc/hdf_convenience.c:2054: error: cast from pointer to
    integer of different size

The problem is that at these two locations, hid_t are being cast to
and from pointers. Now, hid_t is an int on my machine, but on a 64-bit
machine, ints are 32-bits, so they can't hold pointers, which are
64-bits wide. So this casting won't work.

Furthermore, this seems to be breaking the HDF5 API. I don't see in
the API documentation any mention that hid_t can be used as a pointer,
so it's not clear to me that HDF5 itself needs a patch to
conditionally define hid_t according to the architecture. This looks
like an unsafe hack that minc tools are doing.

This all matters to me because I'm trying to debug a segfault with
dcm2mnc, but first I need to be able to cleanly compile minc-tools, if
nothing else, to compile in debug symbols.

Thanks,
- Jordi G. H.

From jordigh at octave.org  Mon May 14 18:20:29 2012
From: jordigh at octave.org (=?UTF-8?Q?Jordi_Guti=C3=A9rrez_Hermoso?=)
Date: Mon, 14 May 2012 18:20:29 -0400
Subject: [MINC-development] Fix for a segfault in dcm2mnc
Message-ID: <CAPHS2gwnNfKhSGAx5pWAm64PajOZZe0Yy_mq1xEQk8SX-4XzzA@mail.gmail.com>

Hi, I have patched on the development branch a problem with an
uninitialised variable that was causing dcm2mnc to segfault. You can
pull my fix here:

    http://inversethought.com/hg/minc-tools/graph

There is another commit there for silencing some warnings that could
be problematic.

HTH,
- Jordi G. H.

From a.janke at gmail.com  Tue May 15 10:07:51 2012
From: a.janke at gmail.com (Andrew Janke)
Date: Wed, 16 May 2012 00:07:51 +1000
Subject: [MINC-development] Fix for a segfault in dcm2mnc
In-Reply-To: <CAPHS2gwnNfKhSGAx5pWAm64PajOZZe0Yy_mq1xEQk8SX-4XzzA@mail.gmail.com>
References: <CAPHS2gwnNfKhSGAx5pWAm64PajOZZe0Yy_mq1xEQk8SX-4XzzA@mail.gmail.com>
Message-ID: <CAG0CB23G1pa3+2amohzCpkruE8v3g9PX+QbedXrBUx-3QDkXnw@mail.gmail.com>

Hi Jordi,

Thanks for the patches, I have made a commit to the develop tree here with them:

   https://github.com/BIC-MNI/minc/commit/9459071d2ccf1c32a92e926f14f612e6262f4c3c

Thanks again,


a



On 15 May 2012 08:20, Jordi Guti?rrez Hermoso <jordigh at octave.org> wrote:
> Hi, I have patched on the development branch a problem with an
> uninitialised variable that was causing dcm2mnc to segfault. You can
> pull my fix here:
>
> ? ?http://inversethought.com/hg/minc-tools/graph
>
> There is another commit there for silencing some warnings that could
> be problematic.
>
> HTH,
> - Jordi G. H.
> _______________________________________________
> MINC-development mailing list
> MINC-development at bic.mni.mcgill.ca
> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-development