#215 gscan2pdf v1.3.6_r7_gd975a97 feedback

[Alistair]

In Bug #213, Jeffrey wrote:

I've pushed all my commits now, so please go ahead and test the fix

Thanks, Jeffrey. From my perspective this is a big improvement on 1.3.6. I'm back to being able to scan multiple pages with rotate, unpaper and OCR.

I did notice a few issues:

1. The progress bar at the bottom isn't behaving as it used to. It's showing the scan progress, but disappears after the scan finishes, even though rotate, unpaper and OCR are being executed (they all complete successfully, just no progress bar).

2. The default save date is 'tomrrow', i.e. today is 12 Jan 2016, the suggested save date is 13 Jan 2016. I'm in Melbourne, Australia, which is currently on DST: GST+11.

This only happens on startup. After it has been changed to 'today' it stays there.

1. After successfully running unpaper, OCR failed with:

WARN - *** unhandled exception in callback:
*** Can't call method "get_active" on an undefined value at /usr/bin/gscan2pdf line 5967.
*** ignoring at /usr/bin/gscan2pdf line 7106.

This seems to be related to the "Encoder did not produce proper pts, making some up" pop-up dialog appearing, but I'm not 100% sure, and it only happened a few times.

I've attached a log, the error messages start on line 53442 (I tried to run OCR three times).

Thanks!
Alistair

[Jeffrey Ratcliffe]

Thanks for spotting the issue with the progress bar. I'll fix that, push the commit and let you know.

The issue with the date I think is a red herring. I haven't touched that code in a couple of release. If you can reproduce the problem reliably, please let me know.

The OCR issue I can't reproduce at all. You said it only happened a few times. Have you any idea what was different those times?

[Alistair]

Thanks for spotting the issue with the progress bar. I'll fix that, push the commit and let you know.

Glad to help... :-)

The issue with the date I think is a red herring. I haven't touched that code in a couple of release. If you can reproduce the problem reliably, please let me know.

I'll investigate further...

The OCR issue I can't reproduce at all. You said it only happened a few times. Have you any idea what was different those times?

I've been able to reproduce this by:

1. Scan a page without rotate, clean up or OCR set.
2. Tools -> OCR

INFO - Importing scan with resolution=600
DEBUG - signal 'finished-process' emitted with data: scan_pages
INFO - Header suggests 34794437
INFO - Expecting 34794437, found 34794437
INFO - New page filename /tmp/gscan2pdf-WQJR/bznDy9RKxi.pnm, format Portable anymap
INFO - Added /tmp/gscan2pdf-WQJR/mloK7N8ohB.pnm at page 1 with resolution 600
DEBUG - Started setting page_number_start from 1 to 2
DEBUG - Started setting max_pages from 0 to -1
DEBUG - Finished setting max_pages from 0 to -1
DEBUG - Finished setting page_number_start from 1 to 2
WARN - *** unhandled exception in callback:
*** Can't call method "get_active" on an undefined value at /usr/bin/gscan2pdf line 5967.
*** ignoring at /usr/bin/gscan2pdf line 7106.

And the dialog doesn't close.

I think the difference is that selecting OCR in the scan dialog works, scanning and then manually running OCR triggers the error.

Thanks!
Alistair

[Jeffrey Ratcliffe]

I've nailed the OCR problem. You didn't have all possible OCR engines installed. I couldn't reproduce it because I had installed all of them. When I tested on a different machine with just tesseract I found the same problem as you.

I've pushed the fixes to that and the progress bar.

I'll hold off closing the bug until you confirm.

[Alistair]

Thanks! The progress bar and OCR issues are both resolved. My memory tells me that I saw the date issue more than once, but I haven't been able to reproduce it again, so my apologies for the trouble.

Just FYI: I did notice the following Gtk critical message in the log, I don't know whether it is important or not, it doesn't seem to affect operation negatively as far as I can tell (although I'm not watching the screen the whole time while scanning):

INFO - Rotating /tmp/gscan2pdf-qrb0/oAnQtiJbzP.pnm by 90 degrees
INFO - Replaced /tmp/gscan2pdf-qrb0/l71jsIv3We.pnm at page 1 with /tmp/gs
can2pdf-qrb0/l71jsIv3We.pnm, resolution 600
INFO - unpaper --black-threshold 0.33 --border-margin 0,0 --deskew-scan-d
irection left,right --layout single --output-pages 1 --white-threshold 0.
9 --overwrite /tmp/gscan2pdf-qrb0/l71jsIv3We.pnm /tmp/gscan2pdf-qrb0/TaMB
rV0XCQ.pnm ;
WARN - Gtk-CRITICAL **: IA__gtk_progress_set_percentage: assertion 'percentage >= 0 && percentage <= 1.0' failed at /usr/bin/gscan2pdf line 1855.

Thanks again,
Alistair

[Jeffrey Ratcliffe]

Good catch. I've just pushed a commit which I think fixes things. Time for a release, methinks.

I'll have another look at your date problem, first though. I have an idea.

[Jeffrey Ratcliffe]

At the moment, the default date is stored in the settings as an offset to today's date. Evidently, there is a bug calculating the offset. This probably only shows itself because you are so close to the International Date Line.

Obviously, I could spend some time looking for the bug, writing a test, and fixing it. However, I don't see the advantage anymore in storing the date as an offset.

How about if I change things simply to store the last date used. The text/date widget makes it easy and quick to change it to anything you like - even dates 1000 years ago.

What do you think?

[Alistair]

At the moment, the default date is stored in the settings as an offset to
today's date. Evidently, there is a bug calculating the offset. This
probably only shows itself because you are so close to the International
Date Line.

Obviously, I could spend some time looking for the bug, writing a test, and
fixing it. However, I don't see the advantage anymore in storing the date as
an offset.

How about if I change things simply to store the last date used. The
text/date widget makes it easy and quick to change it to anything you like -
even dates 1000 years ago.

What do you thing?

I've seen a date problem a couple of times today, but it is slightly different from what I originally described, I selected a date in the save dialog, e.g. 2015-12-04, but when the PDF filename dialog is displayed it shows the day after, i.e. 2015-12-05 (two screenshots attached).

If you do want to change the default date, the two defaults that I can think of are:

1. Toady's date
2. The last date used

I expect that whichever way you go some people are going to be unhappy, unless you proved a preference setting for which to use. I'm OK with either one.

On 15 January 2016 at 08:34, Jeffrey Ratcliffe ra28145@users.sf.net wrote:

Good catch. I've just pushed a commit which I think fixes things. Time for a
release, methinks.

Yep, the Gtk critical messages are gone (gscan2pdf 1.3.6_15_ga29b7f1).

I haven't done extensive testing, just my normal use cases, however even with the date issue, my subjective feeling is that it is in much better shape than 1.3.5 or 1.3.6 - release away! :-)

I'm going to be travelling from Monday for a couple of days and probably busy the rest of the week, so may not be very responsive during that time.

Thanks again,
Alistair

[Jeffrey Ratcliffe]

I've just push a commit, which I hope fixes the date problem.

Would you mind testing it, please?

[Alistair]

Tested - and the problem has gone.

As always, thanks very much for making gscan2pdf available!

Version: gscan2pdf 1.3.8_3_gfd0860d

Cheers,
Alistair