[Pkg-clamav-devel] tomfastmath is ready (take 2)

[Sebastian Andrzej Siewior]

On 2015-11-04 13:39:14 [+0100], Andreas Cadhalpun wrote:
> Hi Sebastian,
Hi Andreas,

> I consider it OK to cast const away, when one has manually checked that
> it's not a problem. But removing the const in clamav should be fine, too.
> 
> >> It would also be useful to check what other programs using tomsfastmath
> >> in Debian expect.
> > Okay, I can do it.
> 
> Anyway, if it's not only clamav that expects the const, patching tomsfastmath
> might still be the better approach.

It is tricky. tomcrypt has a tfm layer where it can be be compiled
against tomfastmath instead of tommath. This one casts the const away [0].
There is a similar function in tommath called mp_read_radix() which is
const [1].
wolfcrypt is the only package that codesearch.d.n found that uses the
function as const. It seems they used tommath in the past and made then
a layer which now invokes tomfastmath in the background and therefore they
kept the const [2].

I will try to address that header change upstream but I doubt there will
be a new release just with this change :) So I would say it is up to you
whether we should ship that patch in tfm or not. I'm fine both ways.

[0] https://github.com/libtom/libtomcrypt/blob/develop/src/math/tfm_desc.c#L171
[1] https://github.com/libtom/libtommath/blob/develop/bn_mp_read_radix.c#L19
[2] http://sources.debian.net/src/wolfssl/3.4.8%2Bdfsg-1/wolfcrypt/src/tfm.c/?hl=2499#L2558

> Best regards,
> Andreas
> 

Sebastian