On Tue, 28 Jan 2014 17:28:16 +0100
Aurelien Jarno <[email protected]> wrote:
> The problem reported by the testsuite is real, and ignoring it means
> that bzip2 compression support is broken on s390x (and other 64-bit
> big-endian architectures).
>
> The problem is that some pointers are casted from long to unsigned int,
> while they do not have the same size. The patch below fixes that using
> a temporary variable as the pointer. There might be better solutions
> like changing the variable type, but I leave that for the upstream.
>
> --- libgringotts-1.2.1.orig/src/libgrg_crypt.c
> +++ libgringotts-1.2.1/src/libgrg_crypt.c
> @@ -347,10 +347,14 @@ decrypt_mem (const GRG_CTX gctx, const G
> }
>
> if (gctx->comp_algo) //bz2
> + {
> + unsigned int uint_oDim = oDim;
> err = BZ2_bzBuffToBuffDecompress ((unsigned char *)
> - tmpData, (unsigned
> int *) &oDim,
> + tmpData, &uint_oDim,
> (unsigned char *)
> curdata, curlen,
> USE_BZ2_SMALL_MEM, 0);
> + oDim = uint_oDim;
> + }
> else //zlib
> err = uncompress (tmpData, &oDim, curdata, curlen);
>
> @@ -411,13 +415,16 @@ grg_encrypt_mem (const GRG_CTX gctx, con
>
> //compress the data
> if (gctx->comp_algo) //bz2
> + {
> + unsigned int uint_compDim = &compDim;
> err = BZ2_bzBuffToBuffCompress (compData,
> - (unsigned int *)
> - &compDim,
> + &uint_compDim,
> (unsigned char *)
> origData, uncDim,
> gctx->comp_lvl * 3, 0,
> 0);
> + compDim = uint_compDim;
> + }
> else
> err = compress2 (compData, &compDim, origData, uncDim,
> gctx->comp_lvl * 3);
>
> --
Thanks Aurelien for your report and patch. I added it to a new version
uploaded to mentors. And some fixes that will close the other RC bug
(736938)[0].
Ah, on line 415 the assignment should be unsigned int uint_compDim =
compDim not his pointer.
I'm sorry regard to the problems this causes. Hope this new revision
fixes all problems. Anyway there are many inconsistencies and I expect
upstream could take a look.
Regards,
[0] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=736938
--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]
