Tim, et al, I have run into several problems with V0.8.0. I will address them seperately. 1. My compiler is complaining about the following code in celt.c which seems to define metric first as celt_word32, then as celt_word16. Am I mis-interpreting something? ?? VARDECL(celt_word32, metric); ?? ALLOC(metric, len, celt_word16); Thx MikeH -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.xiph.org/pipermail/opus/attachments/20100706/ace6cbe4/attachment-0002.htm
Hi Mike, Indeed, you found a bug! I missed it because the symptoms you saw only occurs in fixed-point with compilers that don't have var arrays or alloca(). It's fixed in git now. For now (commit b6a5f9), the git master branch is bitstream compatible with 0.8.0. I'll probably issue a 0.8.1 bugfix release after we fix any remaining problems you have. Cheers, Jean-Marc On 10-07-06 04:52 PM, Mike Hooper Hooper wrote:> Tim, et al, > I have run into several problems with V0.8.0. I will address them > seperately. > 1. My compiler is complaining about the following code in celt.c which > seems to define metric first as celt_word32, then as celt_word16. Am I > mis-interpreting something? > VARDECL(celt_word32, metric); > ALLOC(metric, len, celt_word16); > Thx > MikeH > > > > _______________________________________________ > celt-dev mailing list > celt-dev at xiph.org > http://lists.xiph.org/mailman/listinfo/celt-dev
Jean-Marc, ? I wish I could take credit for finding the bug, but I'll have to give credit to my compiler....:) ? Let's now take a look at this code in bands.c "???? quant_band(encode, m, i, X, Y, N, b, spread, tf_change, lowband, resynth, ec, &remaining_bits, LM, norm+M*eBands[i], bandE, 0);" ? My compiler says: "argument of type "int *" is incompatible with parameter of type "celt_int32 *"????? libcelt80/libceltbands.c???? line 951??? 1278452093812???? 9084" ? I'm not exactly sure where the problem is. Can you see it? ? MikeH ? ? -----Original Message----- From: Jean-Marc Valin [mailto:jean-marc.valin at usherbrooke.ca] Sent: Tuesday, July 06, 2010 6:00 PM To: Mike Hooper Hooper Cc: celt-dev at xiph.org Subject: Re: [CELT-dev] V0.8.0 Problems ? Hi Mike, ? Indeed, you found a bug! I missed it because the symptoms you saw only occurs in fixed-point with compilers that don't have var arrays or alloca(). It's fixed in git now. For now (commit b6a5f9), the git master branch is bitstream compatible with 0.8.0. I'll probably issue a 0.8.1 bugfix release after we fix any remaining problems you have. ? Cheers, ? ????? Jean-Marc ? ? On 10-07-06 04:52 PM, Mike Hooper Hooper wrote:> Tim, et al, > I have run into several problems with V0.8.0. I will address them > seperately. > 1. My compiler is complaining about the following code in celt.c which > seems to define metric first as celt_word32, then as celt_word16. Am I > mis-interpreting something? > VARDECL(celt_word32, metric); > ALLOC(metric, len, celt_word16); > Thx > MikeH >?>?>?> _______________________________________________ > celt-dev mailing list > celt-dev at xiph.org > http://lists.xiph.org/mailman/listinfo/celt-dev-------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.xiph.org/pipermail/opus/attachments/20100706/ed404985/attachment-0002.htm
Hi Mike, Thanks again on that other bug. It turns out that on most platforms "celt_int32" is the same as "int" so the compiler doesn't complain. Unfortunately, I haven't found any way to test for these kinds of issues that happen only on 16-bit platforms and I've had to rely on people like you compiling on such platforms. Let me know if you find anything else. This one is now fixed in git. Thanks, Jean-Marc On 10-07-06 06:22 PM, Mike Hooper Hooper wrote:> Jean-Marc, > > I wish I could take credit for finding the bug, but I'll have to give > credit to my compiler....:) > > Let's now take a look at this code in bands.c > > " quant_band(encode, m, i, X, Y, N, b, spread, tf_change, lowband, > resynth, ec, &remaining_bits, LM, norm+M*eBands[i], bandE, 0);" > > My compiler says: > > "argument of type "int *" is incompatible with parameter of type > "celt_int32 *" libcelt80/libcelt bands.c line 951 1278452093812 9084" > > I'm not exactly sure where the problem is. Can you see it? > > MikeH > > -----Original Message----- > > From: Jean-Marc Valin [mailto:jean-marc.valin at usherbrooke.ca] > > Sent: Tuesday, July 06, 2010 6:00 PM > > To: Mike Hooper Hooper > > Cc: celt-dev at xiph.org > > Subject: Re: [CELT-dev] V0.8.0 Problems > > Hi Mike, > > Indeed, you found a bug! I missed it because the symptoms you saw only > > occurs in fixed-point with compilers that don't have var arrays or > > alloca(). It's fixed in git now. For now (commit b6a5f9), the git master > > branch is bitstream compatible with 0.8.0. I'll probably issue a 0.8.1 > > bugfix release after we fix any remaining problems you have. > > Cheers, > > Jean-Marc > > On 10-07-06 04:52 PM, Mike Hooper Hooper wrote: > > > Tim, et al, > > > I have run into several problems with V0.8.0. I will address them > > > seperately. > > > 1. My compiler is complaining about the following code in celt.c which > > > seems to define metric first as celt_word32, then as celt_word16. Am I > > > mis-interpreting something? > > > VARDECL(celt_word32, metric); > > > ALLOC(metric, len, celt_word16); > > > Thx > > > MikeH > > > > > > > > > > > > _______________________________________________ > > > celt-dev mailing list > > > celt-dev at xiph.org > > > http://lists.xiph.org/mailman/listinfo/celt-dev >