Re: [FFmpeg-devel] [PATCH] aacpsy: avoid norm_fac becoming NaN
On Thu, Apr 16, 2015 at 1:00 PM, Andreas Cadhalpun
andreas.cadhal...@googlemail.com wrote:
If both band-active_lines and band-thr are 0.0f, the division is
undefined, making norm_fac not a number.
NaN is passed on to other variables until it finally reaches
sce-sf_idx and is converted to an integer (-2147483648).
This causes a segmentation fault when it is used as array index.
Signed-off-by: Andreas Cadhalpun andreas.cadhal...@googlemail.com
---
libavcodec/aacpsy.c | 5 -
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/libavcodec/aacpsy.c b/libavcodec/aacpsy.c
index d1e65f6..b71933b 100644
--- a/libavcodec/aacpsy.c
+++ b/libavcodec/aacpsy.c
@@ -727,7 +727,10 @@ static void psy_3gpp_analyze_channel(FFPsyContext *ctx,
int channel,
if (active_lines 0.0f)
band-thr = calc_reduced_thr_3gpp(band,
coeffs[g].min_snr, reduction);
pe += calc_pe_3gpp(band);
-band-norm_fac = band-active_lines / band-thr;
+if (band-active_lines != 0.0f)
+band-norm_fac = band-active_lines / band-thr;
+else
+band-norm_fac = 0.0f;
norm_fac += band-norm_fac;
}
}
It should be if band-thr 0.0f, all divisions by zero return
something that casts into an ~1:
Try:
#include stdio.h
int main() {
printf(%d\n, int(0.0f / 0.0f));
printf(%d\n, int(1.0f / 0.0f));
return 0;
}
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] aacpsy: avoid norm_fac becoming NaN
On Thu, Apr 16, 2015 at 10:41 AM Claudio Freire klaussfre...@gmail.com wrote: It should be if band-thr 0.0f, all divisions by zero return something that casts into an ~1: Isn't division by zero undefined behavior? Timothy ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] aacpsy: avoid norm_fac becoming NaN
On Thu, Apr 16, 2015 at 05:47:50PM +, Timothy Gu wrote: On Thu, Apr 16, 2015 at 10:41 AM Claudio Freire klaussfre...@gmail.com wrote: It should be if band-thr 0.0f, all divisions by zero return something that casts into an ~1: Isn't division by zero undefined behavior? for floats its defined by IEEE754 if the implementation supports that if iam not mistaken [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB The educated differ from the uneducated as much as the living from the dead. -- Aristotle signature.asc Description: Digital signature ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] aacpsy: avoid norm_fac becoming NaN
On Thu, Apr 16, 2015 at 02:41:39PM -0300, Claudio Freire wrote: On Thu, Apr 16, 2015 at 1:00 PM, Andreas Cadhalpun andreas.cadhal...@googlemail.com wrote: If both band-active_lines and band-thr are 0.0f, the division is undefined, making norm_fac not a number. NaN is passed on to other variables until it finally reaches sce-sf_idx and is converted to an integer (-2147483648). This causes a segmentation fault when it is used as array index. Signed-off-by: Andreas Cadhalpun andreas.cadhal...@googlemail.com --- libavcodec/aacpsy.c | 5 - 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/libavcodec/aacpsy.c b/libavcodec/aacpsy.c index d1e65f6..b71933b 100644 --- a/libavcodec/aacpsy.c +++ b/libavcodec/aacpsy.c @@ -727,7 +727,10 @@ static void psy_3gpp_analyze_channel(FFPsyContext *ctx, int channel, if (active_lines 0.0f) band-thr = calc_reduced_thr_3gpp(band, coeffs[g].min_snr, reduction); pe += calc_pe_3gpp(band); -band-norm_fac = band-active_lines / band-thr; +if (band-active_lines != 0.0f) +band-norm_fac = band-active_lines / band-thr; +else +band-norm_fac = 0.0f; norm_fac += band-norm_fac; } } It should be if band-thr 0.0f, ok, thanks all divisions by zero return something that casts into an ~1: casting NaN or any out of int range value to int is undefined AFAIK [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB The real ebay dictionary, page 3 Rare item - Common item with rare defect or maybe just a lie Professional - 'Toy' made in china, not functional except as doorstop Experts will know - The seller hopes you are not an expert signature.asc Description: Digital signature ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] aacpsy: avoid norm_fac becoming NaN
On 16.04.2015 19:41, Claudio Freire wrote: It should be if band-thr 0.0f, all divisions by zero return something that casts into an ~1: OK, new patch attached. If band-thr is 0.0f and band-active_lines is not, it takes a bit longer to get to NaN. I assume band-thr 0.0f is not possible? Best regards, Andreas From e9124e077f1e8fcef0669ceff2fc05116b942b5d Mon Sep 17 00:00:00 2001 From: Andreas Cadhalpun andreas.cadhal...@googlemail.com Date: Thu, 16 Apr 2015 20:04:54 +0200 Subject: [PATCH] aacpsy: avoid psy_band-threshold becoming NaN If band-thr is 0.0f, the division is undefined, making norm_fac not a number or infinity, which causes psy_band-threshold to become NaN. This is passed on to other variables until it finally reaches sce-sf_idx and is converted to an integer (-2147483648). This causes a segmentation fault when it is used as array index. Signed-off-by: Andreas Cadhalpun andreas.cadhal...@googlemail.com --- libavcodec/aacpsy.c | 5 - 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/libavcodec/aacpsy.c b/libavcodec/aacpsy.c index d1e65f6..7205ee3 100644 --- a/libavcodec/aacpsy.c +++ b/libavcodec/aacpsy.c @@ -727,7 +727,10 @@ static void psy_3gpp_analyze_channel(FFPsyContext *ctx, int channel, if (active_lines 0.0f) band-thr = calc_reduced_thr_3gpp(band, coeffs[g].min_snr, reduction); pe += calc_pe_3gpp(band); -band-norm_fac = band-active_lines / band-thr; +if (band-thr 0.0f) +band-norm_fac = band-active_lines / band-thr; +else +band-norm_fac = 0.0f; norm_fac += band-norm_fac; } } -- 2.1.4 ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] aacpsy: avoid norm_fac becoming NaN
On Thu, Apr 16, 2015 at 3:19 PM, Andreas Cadhalpun andreas.cadhal...@googlemail.com wrote: On 16.04.2015 19:41, Claudio Freire wrote: It should be if band-thr 0.0f, all divisions by zero return something that casts into an ~1: OK, new patch attached. If band-thr is 0.0f and band-active_lines is not, it takes a bit longer to get to NaN. I assume band-thr 0.0f is not possible? It's not, it's an absolute energy threshold. LGTM now. ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
Re: [FFmpeg-devel] [PATCH] aacpsy: avoid norm_fac becoming NaN
On Thu, Apr 16, 2015 at 03:26:55PM -0300, Claudio Freire wrote: On Thu, Apr 16, 2015 at 3:19 PM, Andreas Cadhalpun andreas.cadhal...@googlemail.com wrote: On 16.04.2015 19:41, Claudio Freire wrote: It should be if band-thr 0.0f, all divisions by zero return something that casts into an ~1: OK, new patch attached. If band-thr is 0.0f and band-active_lines is not, it takes a bit longer to get to NaN. I assume band-thr 0.0f is not possible? It's not, it's an absolute energy threshold. LGTM now. applied thanks [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB Everything should be made as simple as possible, but not simpler. -- Albert Einstein signature.asc Description: Digital signature ___ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
