Beta testers for calibre 3.0 needed

[eschwartz]

Originally Posted by kovidgoyal

@eschwartz: No can't say I have heard from jelle -- it has always been on my TODO list to unbundle the un-modified stuff once the new calibre build system was in operation.

That is such an interesting coincidence then, because I mentioned it to him on IRC only a couple days ago and you must have psychically picked up on that and decided now is the right time.

hunspell is modified, see commit da7ef93e27

Ah, somehow I failed to check that.

... and looking at the upstream bugtracker, it looks like that is a feature they do want and is scheduled for v2: https://github.com/hunspell/hunspell/pull/443

So hopefully then you can switch back to the system copy.

chardet long predates the upstream chardet (IIRC it came from firefox originally).

It *looks* like someone eventually picked that up and worked off it to create the current https://github.com/chardet/chardet so maybe it is worth switching to that anyway?

tinycss is modified. templite I dont know about (it was used by someone else's code in calibre).

odpy I tried to upstream way back when, without success. Can't say I am particularly inclined to spend time trying again.

Yeah, so basically ignore me. This is what happens when I get "inspired" in the middle of the night right before a flight.

[PandathePanda]

Originally Posted by davidfor

Yes, that works, and I'm sure you have mentioned it earlier in the thread.

Unfortunately, the download links don't work. Tapping them just refreshes the screen. The server log shows the request was received. There must be something about the response the browser doesn't like. Are there more detailed logs I can look at?

Also found this problem. Tried it on UC Browser and it it tried to download the book and failed.

But works (sort of) with Moon+, setup the server and it downloads but won't open it by default, need to tap on the notification to read it.

[Divingduck]

Originally Posted by chaley

Random thought: the error is almost always "max recursion exceeded" when fetching the list of "keys" for a "dictionary" (a type of array indexed by words instead of numbers). One way I can imagine this happening is that calibre is running out of memory and therefore unable to dynamically allocate memory to the "stack". Do you have a lot of memory on this computer? If not, do you have a rather large virtual memory paging file? I have 8GB of memory. My paging file size is controlled by Windows, and is currently 1.2 GB.

One error in the log looked like it happened during translation, when the template formatter was returning an error message. Do the errors go away if you run calibre in English? I tried running in German and everything worked.

Hello Chaley,
Thanks for looking to my problem. Sorry for come back so late. I though I have some time this morning, but was interrupted after I did a new source update and first tests. I have 8 GB memory on this computer too and page file (managed by windows) is 1.25GB. My actual usage in memory is about 2.8 to 3.8 GB

I had this morning two different experiences with calibre:

I download the new source updates this morning at 10:30 and see that all custom columns are now ok except column "Virtuelle Bibliothek" {0:'virtual_libraries()'}, where only the first book in the list is shown correctly and all other books have a template error with "recursion maximum depth exceeded while calling a Python object" and sometimes "This function can be used only in GUI mode" (what I did actual)

When I run the same in debug mode the custom columns have errors again and was not able to switch back to the first status with only column "Virtuelle Bibliothek" errors but sometimes some books have correct filled custom columns and a lot not.

I switch the language between German and English and it seems that with english it is a bit better (but do not count on it, I start calibre 3 times with 3 different results for custom columns.
I too believe this is a memory problem and I vaguely remember, that there was something similar in the past. Maybe I can remember myself some when later

[kovidgoyal]

@davidfor: There aren't any more detailed logs. Basically, the server logs requests, with their response code and if any errors occur.

The only thing I can think of might be the browser doesn't like one of the headers. You can check what headers the server send using either

curl -I

or just using the dev tools in a desktop browser.

[kovidgoyal]

@eschwartz: No, the motivation for me doing regex was this: https://bitbucket.org/mrabarnett/mra...locale-failure

I have to say I love the maintainer of regex. Alongwith apsw they are the only two upstream dependencies of calibre that have actually competent/active maintainers.

I dont mind in principle stripping out chardet. I just need to find the time to do some testing to make sure APIs haven't changed.

[davidfor]

Originally Posted by kovidgoyal

@davidfor: There aren't any more detailed logs. Basically, the server logs requests, with their response code and if any errors occur.

The only thing I can think of might be the browser doesn't like one of the headers. You can check what headers the server send using either

curl -I

or just using the dev tools in a desktop browser.

Comparing what I get with 2.99.8 and portable 1.48, which does work:

E:\Download\Windows\Network\curl-7.34.0-ssl-sspi-zlib-static-bin-w32\curl>curl -I http://10.1.225.111/get/EPUB/939/Unclean
HTTP/1.1 200 OK
Accept-Ranges: bytes
Content-Disposition: attachment; filename="Evil Librarian - Michelle Knudsen_939.epub"
Content-Length: 679272
Content-Type: application/epub+zip
Date: Sun, 28 May 2017 12:06:28 GMT
ETag: "af8fb9c8593adf88c34b735f5c099063ea088426"
Keep-Alive: timeout=120
Server: calibre 2.99.8

E:\Download\Windows\Network\curl-7.34.0-ssl-sspi-zlib-static-bin-w32\curl>curl -I http://10.1.225.111:8080/get/epub/Calibre%20Quick%20Start%20Guide%20-%20John%20Schember_1.epub
HTTP/1.1 200 OK
Content-Length: 151399
Content-Disposition: attachment; filename="Calibre Quick Start Guide - John Schember_1.epub"
Server: calibre/1.48.0
Last-Modified: Tue, 03 Jan 2017 07:07:47 GMT
Date: Sun, 28 May 2017 12:09:09 GMT
Content-Type: application/epub+zip

The extra headers for 2.99.8 is "Accept-Ranges: bytes", "Keep-Alive" and the "ETag". I'll have a play and see what I can come up with.

[eschwartz]

Originally Posted by kovidgoyal

@eschwartz: No, the motivation for me doing regex was this: https://bitbucket.org/mrabarnett/mra...locale-failure

I have to say I love the maintainer of regex. Alongwith apsw they are the only two upstream dependencies of calibre that have actually competent/active maintainers.

Indeed.

I dont mind in principle stripping out chardet. I just need to find the time to do some testing to make sure APIs haven't changed.

Cool, I'll leave you to it.

[m.s]

Originally Posted by davidfor

Unfortunately, the download links don't work. Tapping them just refreshes the screen.

The download don't work only on Kobo-devices (Kobo Glo, Glo HD, Aura One). On other readers (Bookeen, Kindle, PocketBook, Tolino) it works. Maybe the Kobobrowser can´t understand the redirect to the EPUB-file?

[m.s]

Originally Posted by kovidgoyal

Just go to

/mobile

for a simple page.

Could you set a link to the simple page under the "Loading, please wait"-text?

[davidfor]

Originally Posted by m.s

The download don't work only on Kobo-devices (Kobo Glo, Glo HD, Aura One). On other readers (Bookeen, Kindle, PocketBook, Tolino) it works. Maybe the Kobobrowser can´t understand the redirect to the EPUB-file?

It doesn't seem to be that. The request and response look OK. Unfortunately, I can't see what is happening on the device.

[kovidgoyal]

@davidfor: My guess would be Accept-Ranges. You can turn it off easily to test by changing line 676 in http_response.py

@ms: Sure it is easy to add a link to the simple HTML page.

[m.s]

Originally Posted by kovidgoyal

My guess would be Accept-Ranges. You can turn it off easily to test by changing line 676 in http_response.py

Is there a way to test this on calibre portable?

[kovidgoyal]

You just need to set it up to run from source: see the setting up a calibre development environment section of the user manual. Note that if you do, the first time you run the server it will take a ery long time to start, because it will need to compile the RapydSCript code before running.

[m.s]

Thanks. I´ve tested it and without Accept-Ranges it doesn't work too.

[Martina Schein]

Kovid, when uploading the new installation file »calibre-64bit-2.99.9.msi« something has gone wrong. It is much smaller than the previous versions and can not be installed.
The second screenshots illustrate this.