Reports crashes in 0.9.10 - Page 2

DiapDealer · 09-27-2018, 06:52 PM

Sounds good to me. Thanks!

KevinH · 09-28-2018, 09:57 AM

Okay, I just pushed a change to our github master so that HTMLResource objects will use the QXmlStreamReader itself for well-formed checks instead of the earlier gumbo version (which still exists). I rebuilt with these changes and it fixes the issue by preventing Reports from launching if the xhtml in the book is not well-formed.

That said, the WellFormed checks are used in a number of places so we really need to test this change to make sure that increasing strictness here does not create other issues with Sigil being able to handle very poor html as input.

I have my fingers crossed that there will be no side effects!

KevinH

DiapDealer · 09-28-2018, 11:08 AM

Quote:

Originally Posted by KevinH

Okay, I just pushed a change to our github master so that HTMLResource objects will use the QXmlStreamReader itself for well-formed checks instead of the earlier gumbo version (which still exists). I rebuilt with these changes and it fixes the issue by preventing Reports from launching if the xhtml in the book is not well-formed.

That said, the WellFormed checks are used in a number of places so we really need to test this change to make sure that increasing strictness here does not create other issues with Sigil being able to handle very poor html as input.

I have my fingers crossed that there will be no side effects!

KevinH

I grabbed the Windows Appveyor build of your latest commit, and it definitely seems to work for the specific test case I mentioned. What areas/features should I focus on to check for problems, do you think? Splitting/merging relies on things being well-formed, right?

KevinH · 09-28-2018, 11:21 AM

yes, splitting and merging as well as initial reading in of the epub if Mend on open is disabled, and checking each file after a plugin runs are all places we do well-formed checks.

Since strictness increased, I wanted to make sure the reading in of epubs and raw html that may not be well-formed is not prevented or hurt.

Thanks,

Kevin

ghmerrill · 10-01-2018, 02:31 PM

Sorry I haven't been keep up with this for the past few days. As a (previously long time -- Bell Labs, SAS Institute, etc.) compiler writer) I would favor the more formally correct approach that I think you're suggesting -- which I think would also result in minimal intervention in your code base, less of a maintenance headache, and more stability. So both a theoretical and practical better approach.

But in any event, thanks for your prompt attention and very professional response to this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
"date" reports correct time zone and time; UI reports a time several hours off	Synthead	Kindle Developer's Corner	17	01-05-2015 05:10 AM
Calibre crashes on dupes reports page after importing new books	juneloon	Library Management	0	03-09-2014 12:14 AM
Problem with reports on 0.7.4	JimmyG	Sigil	4	11-01-2013 12:55 PM
Calibre reports?	calvin-c	Calibre	1	02-25-2013 09:04 PM
CES Reports	daffy4u	News	0	01-08-2010 10:23 AM

09-27-2018, 06:52 PM	#16
DiapDealer Grand Sorcerer Posts: 27,546 Karma: 193191846 Join Date: Jan 2010 Device: Nexus 7, Kindle Fire HD	Sounds good to me. Thanks!

09-28-2018, 09:57 AM	#17
KevinH Sigil Developer Posts: 7,636 Karma: 5433388 Join Date: Nov 2009 Device: many	Okay, I just pushed a change to our github master so that HTMLResource objects will use the QXmlStreamReader itself for well-formed checks instead of the earlier gumbo version (which still exists). I rebuilt with these changes and it fixes the issue by preventing Reports from launching if the xhtml in the book is not well-formed. That said, the WellFormed checks are used in a number of places so we really need to test this change to make sure that increasing strictness here does not create other issues with Sigil being able to handle very poor html as input. I have my fingers crossed that there will be no side effects! KevinH

09-28-2018, 11:21 AM	#19
KevinH Sigil Developer Posts: 7,636 Karma: 5433388 Join Date: Nov 2009 Device: many	yes, splitting and merging as well as initial reading in of the epub if Mend on open is disabled, and checking each file after a plugin runs are all places we do well-formed checks. Since strictness increased, I wanted to make sure the reading in of epubs and raw html that may not be well-formed is not prevented or hurt. Thanks, Kevin

10-01-2018, 02:31 PM	#20
ghmerrill Connoisseur Posts: 72 Karma: 10 Join Date: Mar 2017 Location: Central NC Device: Various	Sorry I haven't been keep up with this for the past few days. As a (previously long time -- Bell Labs, SAS Institute, etc.) compiler writer) I would favor the more formally correct approach that I think you're suggesting -- which I think would also result in minimal intervention in your code base, less of a maintenance headache, and more stability. So both a theoretical and practical better approach. But in any event, thanks for your prompt attention and very professional response to this.

Advert

Advert