BD Rebuilder Beta - Bug Reports Only

[jdobbs]

Good hint ... and guess what?



That worked!
I set decimal separator to "." and digit grouping to "," (like here (https://www.officetooltips.com/office_2016/tips/how_to_change_decimal_symbol_and_digit_grouping_symbol_in_windows_10.html) ) and now the encode runs fine in CQM mode...

But I'd rather have my settings back to normal :scared:Try this version and let me know if it fixes your issue:

BD-RB v0.61.07 (https://jammernhilftnichts.de/jdobbs/BD-RBV06107.zip)

It also adds the --repeat-headers fix when using UHD_V3_MODE=1

Also, I did some testing yesterday to compare CQM and VBR modes with NVENCC. At the same files size (bitrate) it seems that CQM gives a better SSIM. But... since the CQM prediction isn't perfect, its possible that you might have a lower file size and the better SSIM might not hold true. I'm looking at possible ways to improve prediction, or, since it's so fast -- maybe do a second CQM pass (using the first pass as the ultimate "prediction" sample).

[Sharc]

Also, I did some testing yesterday to compare CQM and VBR modes with NVENCC. At the same files size (bitrate) it seems that CQM gives a better SSIM.
Some time ago I found to my surprise that leaving --aq 'off' gave the best SSIM and VMAF values in CQM mode (adjusted for nearly same filesize) using NVEncC (AVC, h.264). Maybe it was specific to the source I tested and/or the --aq algo may have improved since, so don't quote me on that.....

[GreasyGreedo]

Have you had a chance to look into the issue where subtitles are lost from a UHD disc containing multiple movie files? Just checking, no rush or anything.

[Michi]

CQM fails on 61.07, too:

[07.19.20] BD Rebuilder v0.61.07
[19:50:44] Source: UHD_GENTLEMEN
- Input BD size: 60,94 GB
- Approximate total content: [01:56:59.095]
- Target BD size: 46,26 GB
- Windows Version: 6.2 [9200]
- Quality: Highest (Very Slow), CQM
- HEVC/V3 mode for BD disc enabled
- Decoding/Frame serving: NVENCC
- Audio Settings: AC3=0 DTS=0 HD=1 Kbs=448
[19:50:46] PHASE ONE, Encoding
- [19:50:46] Processing: VID_00000 (1 of 2)
- [19:50:46] Extracting A/V streams [VID_00000]
- [19:50:50] Reencoding video [VID_00000]
- Source Video: HEVC, 3840x2160
- Rate/Length: 23,976fps, 610 frames
- [19:50:50] Performing CQM Prediction...
- Analyzing 1,00 [19:51:01] - Failed video encode, aborted


Nvenc VBR works without error, but the quality on UHDs with grain isn't good enough for me.

I used a GTX 1660 for all tests.

[Sharc]

CQM fails on 61.07, too:


- Source Video: HEVC, 3840x2160
- Rate/Length: 23,976fps, 610 frames
- [19:50:50] Performing CQM Prediction...
- Analyzing 1,00 [19:51:01] - Failed video encode, aborted


The blue seems not right to me ....
If size is not that important to you, you may want to try with a fix quality factor rather than running the prediction, or you could try with the old prediction algo perhaps (NEW_CRF_PREDICTION=0 in the .ini). I didn't try it though.

[jdobbs]

CQM fails on 61.07, too:

[07.19.20] BD Rebuilder v0.61.07
[19:50:44] Source: UHD_GENTLEMEN
- Input BD size: 60,94 GB
- Approximate total content: [01:56:59.095]
- Target BD size: 46,26 GB
- Windows Version: 6.2 [9200]
- Quality: Highest (Very Slow), CQM
- HEVC/V3 mode for BD disc enabled
- Decoding/Frame serving: NVENCC
- Audio Settings: AC3=0 DTS=0 HD=1 Kbs=448
[19:50:46] PHASE ONE, Encoding
- [19:50:46] Processing: VID_00000 (1 of 2)
- [19:50:46] Extracting A/V streams [VID_00000]
- [19:50:50] Reencoding video [VID_00000]
- Source Video: HEVC, 3840x2160
- Rate/Length: 23,976fps, 610 frames
- [19:50:50] Performing CQM Prediction...
- Analyzing 1,00 [19:51:01] - Failed video encode, aborted


Nvenc VBR works without error, but the quality on UHDs with grain isn't good enough for me.

I used a GTX 1660 for all tests.I'll take another look at it.

Can you post the contents of LASTCMD.TXT (in the BD-RB folder) right after it fails? If possible, then copy the command to a DOS window and try and run it so we can see what error is occurring.

[Michi]

LASTCMD.TXT;

"C:\BD_Rebuilder UHD\tools\nvenc\nvencc.exe" --avhw -i "Q:\4K\WORKFILES\00000.AVS.SMPL.m2ts" --codec hevc --preset
quality --profile main10 --output-depth 10 --repeat-headers --chromaloc 2 --colorprim bt2020 --transfer smpte2084
--colormatrix bt2020nc --master-display G(13250,34500)B(7500,3000)R(34000,16000)WP(15635,16450)L(10000000,1)
--qp-min 0 --vbr 0 --vbr-quality 1 --aq-temporal --sar 1:1 --aud --pic-struct --vbv-bufsize 60000 --max-bitrate 60000
--gop-len 24 -o "Q:\4K\WORKFILES\TEMP.265"

[jdobbs]

@Michi

Thanks. It looks exactly like what cartman0208 saw. I'm going through sources trying to find one that causes it for me so I can point out exactly why the sample is created with a zero length.

[cartman0208]

I reseted my regionals back to normal and the Prediction fails on 61.07...
[23:03:26] PHASE ONE, Encoding
- [23:03:26] Processing: VID_00000 (1 of 14)
- [23:03:26] Extracting A/V streams [VID_00000]
- [23:03:31] Reencoding video [VID_00000]
- Source Video: HEVC, 3840x2160
- Rate/Length: 24,000fps, 1.046 frames
- [23:03:31] Performing CQM Prediction...
- Analyzing 17,35 [23:03:35] - Failed video encode, aborted


Set it to "US" type:
[23:06:53] PHASE ONE, Encoding
- [23:06:53] Processing: VID_00000 (1 of 14)
- [23:06:53] Extracting A/V streams [VID_00000]
- [23:06:57] Reencoding video [VID_00000]
- Source Video: HEVC, 3840x2160
- Rate/Length: 24.000fps, 1,046 frames
- [23:06:57] Performing CQM Prediction...
- Analyzing 17.35 9.18 5.09 3.04 2.02 1.51 1.26 1.13 1.06 1.03 1.02 1.01 [1.01]
- [23:07:31] Encoding using constant quality mode.


Unfortunately thats not all ... on a later vid of that BD the prediction fails with an error Popup
[22:42:06] PHASE ONE, Encoding
- [22:42:06] Processing: VID_00020 (6 of 14)
- [22:42:06] Reencoding video [VID_00020]
- Source Video: HEVC, 3840x2160
- Rate/Length: 24,000fps, 240 frames
- [22:42:06] Performing CQM Prediction...
- [22:42:06] PredictAndEncode() 00006 2810
[22:42:17] - Aborted by user request

LASTCMD.TXT still holds the command from the previous vid.

What I noticed ... VID_00020 only shows the copyright warning, still frame with no motion at all

[jdobbs]

Yeah. Definitely some work to be done on that HEVC prediction routine.

[Moviewatcher666]

I've created a test release (v0.61.06) of BD Rebuilder for anyone who would like to experiment and report any issues they see with the new support for NVIDIA/NVENC support.

Please note that in order to use the new capability you must run BD-RB on a system that has an NVIDIA card installed that supports NVENC. You can find out more about which cards can be used at this link (https://developer.nvidia.com/video-encode-decode-gpu-support-matrix). BD-RB does a check when you attempt to select the NVENC encoder (under the SETTINGS menu).

There were a considerable number of changes to this release, so this version should be considered "for testing purposes only", so I'm not updating the first post of this thread -- please try it and let me know your experiences and report any bugs you find.

BD-RB v0.61.06 (https://jammernhilftnichts.de/jdobbs/BD-RBV06106.zip)

Thanks.

HOLY CRAP! Are you kidding me??

Just did an whole disk 4K encode on my Empire Strike Back disk.

Previous whole disk encodes shrinking to BD50 using FFMPEG frame serving would take 2 days (that's right, 48 HOURS).

Using this version of BD-RB using NVENC frame serving....2 HOURS! And the quality is amazing on my LG 65" OLED TV.

Can there be that much of a difference!? I am truly blown away at the speed difference.

Good work, jdobbs and thank you!

[jdobbs]

HOLY CRAP! Are you kidding me??

Just did an whole disk 4K encode on my Empire Strike Back disk.

Previous whole disk encodes shrinking to BD50 using FFMPEG frame serving would take 2 days (that's right, 48 HOURS).

Using this version of BD-RB using NVENC frame serving....2 HOURS! And the quality is amazing on my LG 65" OLED TV.

Can there be that much of a difference!? I am truly blown away at the speed difference.

Good work, jdobbs and thank you!Good to hear.

[musiclover]

Yeah. Definitely some work to be done on that HEVC prediction routine.

I guess the AVC prediction routine also needs work. With software compression BDRB uses the same bitrate as with hardware compression. But NVENCC compresses the video much more than x264 does. Therefore the resulting BD's are much smaller. So I think NVENCC needs to use a higher bitrate.

But maybe at the moment most of the work must be done on making UHD work properly

[geekzapoppin]

Just tried out the latest beta version using NVENC as the encoder. As a previous post mentioned, it's like BD-Rebuilder got a nitrous injection. I didn't encounter any errors and the results are as good as I've come to expect over the years. Great work!

[cartman0208]

It also adds the --repeat-headers fix when using UHD_V3_MODE=1


First run looks good, no issues.
So basically we get an UHD compliant (HEVC, 10bit) disc with 1080p from a standard BD ... NICE! :D

Can't wait to compare the results on the big screen...

[geekzapoppin]

Spoke a little too soon, I'm afraid. I'm running into a situation where HD audio is not converting to AC3 properly. I don't encounter any errors during the process, but when I test out the final result, there is no audio. I took a look into the WORKFILES folder and noticed that, before the remuxing stage, the AC3 files are 1KB. The intermediate files look to be the correct size, it's like it just gets to the AC3 encoding stage and then either skips encoding or starts long enough to create a tiny dummy file, but then moves on to remuxing when there isn't an actual AC3 file to include. I'm using v0.61.06, but I noticed this happening on older versions as well.

What kind of info do I need to post that would be helpful to you? Just let me know. Your work over the years has been much-appreciated.

[musiclover]

Spoke a little too soon, I'm afraid. I'm running into a situation where HD audio is not converting to AC3 properly. I don't encounter any errors during the process, but when I test out the final result, there is no audio. I took a look into the WORKFILES folder and noticed that, before the remuxing stage, the AC3 files are 1KB. The intermediate files look to be the correct size, it's like it just gets to the AC3 encoding stage and then either skips encoding or starts long enough to create a tiny dummy file, but then moves on to remuxing when there isn't an actual AC3 file to include. I'm using v0.61.06, but I noticed this happening on older versions as well.

What kind of info do I need to post that would be helpful to you? Just let me know. Your work over the years has been much-appreciated.

You need to install the helper apps from the first page again. That usually helps me in this situation

[geekzapoppin]

You need to install the helper apps from the first page again. That usually helps me in this situation

I'll give that a try. Thanks!

[MrVideo]

But I'd rather have my settings back to normal :scared:
Normal is in the eye of the beholder. :D

[jdobbs]

Update: Still working on the CQM/CRF prediction for HEVC sources. I think I've fixed the error issues -- but I'm now trying to try and make it a little more accurate -- and to add an automatic second pass if it oversizes.

[AmigaFuture]

Wow, I've never noticed that file ... :thanks:

What? Okay, reading through the thread.. I'm a _little_ surprised. The current supplier of the BD-RB files..missed that. Smiling.

[cartman0208]

What? Okay, reading through the thread.. I'm a _little_ surprised. The current supplier of the BD-RB files..missed that. Smiling.

Well, true :o
I never used the Alternate Output feature much, there are other solutions, which are actually a wee bit handier ... but I do now, because of NVENC :rolleyes:

And I have a question: Is it possible to burn in subtitles in Alternate Output?
Because with Plex the whole movie gets reencoded, when you turn on subtitles ... and the Atom Processor of the NAS then has some trouble on 4K sources :scared:

I tried "Turn Subtitle ON" but it's not burnt in...

[Sharc]

Well, true :o
I never used the Alternate Output feature much, there are other solutions, which are actually a wee bit handier ... but I do now, because of NVENC :rolleyes:

And I have a question: Is it possible to burn in subtitles in Alternate Output?
Because with Plex the whole movie gets reencoded, when you turn on subtitles ... and the Atom Processor of the NAS then has some trouble on 4K sources :scared:

I tried "Turn Subtitle ON" but it's not burnt in...
Instead of burning you could try with .srt subtitles which you can download free from opensubtitles.org, for example. Text subtitles are widely supported.

[cartman0208]

Instead of burning you could try with .srt subtitles which you can download free from opensubtitles.org, for example. Text subtitles are widely supported.

That might work, but I'd have to maintain the extra subtitle files
plus...
The Subtitles I'm after (foreign language) are not always available

BUT ... I think I found the answer
In the Setup options I found "Use SupTitle for Alternate Forced Subs"
That transforms the selectable Subtitle into something that Plex doesn't need to transcode (Text?) ... and its still selectable, thats even better than burnt in... :)

[ggtop]

thats even better than burnt in... :)
SupTitle is an AviSynth plugin and burn-in is exactely what it's supposed to do :D

ggtop

[cartman0208]

SupTitle is an AviSynth plugin and burn-in is exactely what it's supposed to do :D

ggtop

Burn-in, as I understand it, is a subtitle encoded in the video stream

But the resulting MKV has no burned-in subtitles ... I can switch between different languages

Never mind ... subtitles are fine for me :)

[ggtop]

Burn-in, as I understand it, is a subtitle encoded in the video stream

YES


But the resulting MKV has no burned-in subtitles ... I can switch between different languages

Never mind ... subtitles are fine for me :)

If subs are fine I suggest to uncheck "Use Suptitle..." again as it will burn-in subs into the encoded videostream. Look into the AVS file(s) holding the AviSynth script. By default BD-RB searches for forced subs having language codes defined in your .INI in that order. Example:
FORCED_SUB_LANG=deu;ger;
As I don't use MKV I'm not 100% it works the same way like it does for MP4, but I'm pretty sure it does. BD-RB muxes subs into MKV container. That's why they are present for you using PLEX are other video players. In any case "Use Suptitle..." is not what you want.
There are a lot of sources without forced subs. In case you can enable sub streams by right-clicking them in the streams' pane. Very likely your current source is such an example.

ggtop

[cartman0208]

Thanks for the explanation, that helps a lot.
And as long as Plex can play subtitles without reencoding, any method is fine for me ;)

@jdobbs: you might want to reconsider copying the color information ... there are a few UHD-BDs where (some) extras are in BT.709 ...

[jdobbs]

Thanks for the explanation, that helps a lot.
And as long as Plex can play subtitles without reencoding, any method is fine for me ;)

@jdobbs: you might want to reconsider copying the color information ... there are a few UHD-BDs where (some) extras are in BT.709 ...It should be keeping the current configuration -- including BT.709. Are you saying it isn't?

[jdobbs]

Okay, everyone, let's try this release and see if it gets rid of the errors. It should also do better prediction (although it can never be perfect). It will also do a self-correcting second CQM pass if the output is too large (you may see that a lot on very small files).

BD-RB v0.61.08 (https://jammernhilftnichts.de/jdobbs/BD-RBV06108.zip)

Please let me know what you find and provide any suggestions you think might make it better.

[cartman0208]

It should be keeping the current configuration -- including BT.709. Are you saying it isn't?

Source files are BT.709 with 4K resolution, after BDRB conversion it's BT.2020

[cartman0208]

Unfortunately the issue with european punctuation still exists

- [20:02:14] Reencoding video [VID_00000]
- Source Video: HEVC, 3840x2160
- Rate/Length: 23,976fps, 128.721 frames
- [20:02:14] Performing CQM Prediction...
- Analyzing 1,00 [20:02:23] - Failed video encode, aborted

Changeing punctuation works

- [20:10:04] Reencoding video [VID_00000]
- Source Video: HEVC, 3840x2160
- Rate/Length: 23.976fps, 128,721 frames
- [20:10:04] Performing CQM Prediction...
- Analyzing 1.00 [1.00]
- [20:10:53] Encoding using constant quality mode.

[Sharc]

@jdobbs/cartman:
Just wondering: Is the initial value of 1.0 for the prediction intentional or is it possibly a bug?
It looks like the 128,721 frames are interpreted as 128 frames (a thousands separator issue, I think). Hence it selects the rate factor 1.0.

Edit: Here the frame numbers show with the usual high comma like 175'243 frames and the initial rate factor is 14.5. All ok......

[cartman0208]

@jdobbs/cartman:
Just wondering: Is the initial value of 1.0 for the prediction intentional or is it possibly a bug?
It looks like the 128,721 frames are interpreted as 128 frames (a thousands separator issue, I think). Hence it selects the rate factor 1.0.

Edit: Here the frame numbers show with the usual high comma like 175'243 frames and the initial rate factor is 14.5. All ok......

Good thinking ... If I set both the 1000s separator and decimal separator to "." (which usually is nowhere the case) the logging in the GUI stops completely, where the separator should be
- [21:58:02] Reencoding video [VID_00000]
- Source Video: HEVC, 3840x2160
- Rate/Length: 23.976fps, 128

BDREBUILDER.LOG:
- [21:58:02] Reencoding video [VID_00000]
- Source Video: HEVC, 3840x2160
- Rate/Length: 23.976fps, 128 721 frames
- [21:58:02] Performing CQM Prediction...
- Analyzing 1.00 [1.00]
- [21:58:51] Encoding using constant quality mode.

But even when the separators are set to " ' " and "." ... the initial (and only) number is 1.00 ... I will check the output size

[Sharc]

Something is odd with your settings.
The thousands grouping is a space char rather than ' in your system.
The decimal separator looks ok.
Hence BD RB still assumes a very short file and sets the ratefactor to 1.0 accordingly.
Try to change the thousands separator in your local windows settings.

[jdobbs]

Source files are BT.709 with 4K resolution, after BDRB conversion it's BT.2020Are they being imported, or are they from a UHD disc?

[jdobbs]

Unfortunately the issue with european punctuation still exists



Changeing punctuation worksI am really confused. BD-RB scans the string and replaces any "," with a "."


I'll look at it again.

Can you look at the command line (LASTCMD.TXT) and see if the "," is still there?

[jdobbs]

Something is odd with your settings.
The thousands grouping is a space char rather than ' in your system.
The decimal separator looks ok.
Hence BD RB still assumes a very short file and sets the ratefactor to 1.0 accordingly.
Try to change the thousands separator in your local windows settings.Is there a standard somewhere in which a ' is used instead of . or , in the thousands separator? I don't think I was aware of that.

So 1 million could be: 1,000,000.00 or 1.000.000,00 or 1'000'000,00?

[Sharc]

I am really confused. BD-RB scans the string and replaces any "," with a "."


I'll look at it again.

Can you look at the command line (LASTCMD.TXT) and see if the "," is still there?
I think cartman's problem is the 1000s grouping character (which should be '), not the decimal point.

[Sharc]

Is there a standard somewhere in which a ' is used instead of . or , in the thousands separator? I don't think I was aware of that.

So 1 million could be: 1,000,000.00 or 1.000.000,00 or 1'000'000,00?
I think the notation is a local (national) habit rather than a strict standard. It can therefore be customized in Windows how the numbers should be presented.

[jdobbs]

I think the notation is a local (national) habit rather than a strict standard. It can therefore be customized in Windows how the numbers should be presented.Still confused as to how that value could be misinterpreted. BD-RB calculates the framecount. I can understand how a string being interpreted as a number could be an issue, or how a number being represented on the screen could show the wrong values because of local representations. But how can a number be misinterpreted when is doesn't change from a number (the actual framecount used in calculations)???

I guess I'm going to have to change my system to a European style locality setting and trace it to find out how this happens.

[Sharc]

Still confused as to how that value could be misinterpreted. BD-RB calculates the framecount. I can understand how a string being interpreted as a number could be an issue, or how a number being represented on the screen could show the wrong values because of local representations. But how can a number be misinterpreted when is doesn't change from a number (the actual framecount used in calculations)???

I guess I'm going to have to change my system to a European style locality setting and trace it to find out how this happens.
No idea. Somewhere happening a conversion numbers <-> strings?
Windows secrets?

[jdobbs]

I just set my system so that the thousands separator is . and the decimal is , -- and it seems to be working properly. So I'm not sure why cartman0208 is getting an error.

@cartman0208

Can you give me some hints on settings that cause an issue? Ignore what is showing on the screen -- that doesn't matter right now. Are you getting errors when the encode is running? Can you show me the command line being used?

The new NVENC subroutine is structured after the original ones that did X265 encodes. I'm confused why it would be acting differently.

[Edit] I have to correct myself. There is an error -- I just realized the sample file was empty -- and OMG I was using a special system format call to create the string used for a seek command that tells it where to start! I must have been in lala land when I wrote that code. And why was it so hard for me to see when I reviewed it 10 times?

v061.09 should be coming in a little while. I just want to test it with the regional settings first.

[cartman0208]

Are they being imported, or are they from a UHD disc?

I'll upload a sample ... could take a while

[cartman0208]

This is default in Germany
https://jammernhilftnichts.de/download/format.png

lastcmd is
"U:\BD_Rebuilder\tools\nvenc\nvencc.exe" --avhw -i "U:\FULLDISC\LUCYUHD\BDMV\STREAM\00000.m2ts" --codec hevc --preset quality --profile main10 --output-depth 10 --repeat-headers --chromaloc 2 --colorprim bt2020 --transfer smpte2084 --colormatrix bt2020nc --master-display G(13250,34500)B(7500,3000)R(34000,16000)WP(15635,16450)L(10000000,50) --max-cll 1000,400 --qp-min 0 --vbr 0 --vbr-quality 1 --aq-temporal --keyfile "U:\BD-OUTPUT\WORKFILES\VID_00000.CHP" --sar 1:1 --aud --pic-struct --vbv-bufsize 45000 --max-bitrate 48000 --gop-len 24 -o "U:\BD-OUTPUT\WORKFILES\VID_00000.AVS.hevc"

[cartman0208]

[Edit] I have to correct myself. There is an error -- I just realized the sample file was empty -- and OMG I was using a special system format call to create the string used for a seek command that tells it where to start! I must have been in lala land when I wrote that code. And why was it so hard for me to see when I reviewed it 10 times?

v061.09 should be coming in a little while. I just want to test it with the regional settings first.

Good thing you sorted it out... I was beginning to believe that I messed up my system :D

[jdobbs]

I'm hoping this version fixes the regional issues reported:

BD-RB v0.61.09 (https://jammernhilftnichts.de/jdobbs/BD-RBV06109.zip)

@cartman0208

Can you confirm this is working on your system? Thanks.

Meanwhile I'm going to look at the BT.709 thing and how hard it would be to include HDR10+ in a future version.

One question: Was the BT.709 source an imported file, or was it BT.709 on an original UHD disc? That way I can know if it is the reencoding or the import that causes the mixup.

[cartman0208]

One question: Was the BT.709 source an imported file, or was it BT.709 on an original UHD disc? That way I can know if it is the reencoding or the import that causes the mixup.

It's an original BD rip... the main movie is fine, also most of the extras

[Edit]
I did two rebuilds and both ran fine, but I still only get - Analyzing 1,00 [1,00] in the log .. at least for the main movie ... is that intended?
[Edit2]
Ok, the former applies only on some sources

[CraigWally]

I'm loving the speed gains on this latest version,
However
Is the "Quicker encode for extras" still working right

I've just tried to The Great Escape BluRay 1080,
and the extras have come out larger than the original with the main movie being compressed more than necessary

I'm just doing some further test, I'll post settings/ini etc later if necessary

EDIT: Hmmm, somethings not right here
If I choose "movie and menus / auto blank extras" - it chooses all the extras, rather than blanking them - seems to be just this disc?!

[jdobbs]

It's an original BD rip... the main movie is fine, also most of the extras

[Edit]
I did two rebuilds and both ran fine, but I still only get - Analyzing 1,00 [1,00] in the log .. at least for the main movie ... is that intended?
[Edit2]
Ok, the former applies only on some sourcesOn the one where you get "Analyzing 1,00", can you look at the sample file and see its size? It would be named after the M2TS being sampled, so if the M2TS was 00058 it would be named 00058.AVS.SMPL.m2ts.

I would think the only time you would see a 1,00 would be when the source file was already extremely small. That would be unusual for a feature, you might see it for an extra (like something that hold several still pictures). I've run numerous jobs and I don't think I've ever seen it. It makes me think the sample file still has a length of zero.

Could you post the log from that job? Also, could you post the contents of the INF file for that job?