Urgent: Loss of page content
I have now seen this happen on three different pages.
The page will look fine in read mode, but then invoke edit, and immediately the page is truncated at some point, with the rest of the page below that missing -- possibly a large amount.
Quitting the editor at that stage, ( without saving) results in display of the truncated page, the rest is missing.
It appears that data in your database is messed up at that point, because there's no recent version in history corresponding to the page in its final (non-truncated) state, which clearly there should be if that page once displayed properly.
Some more clues:
All of the three times, the page was truncated in a table (ie: table on the wiki page). Indeed, it was truncated in the middle of a row, with the result that that tr and table wasn't closed. So, in view mode, that resulted in the boilerplate rest of the page being messed up too, though that's merely a side effect of the not-properly-closed tags.
But it does indicate that this is an error that would normally not occur, even by user mistake, since I believe your editor save code goes to some lengths to make sure tags are closed. (I tried to manually delete closing tags in raw html mode, then save, and the page still displayed plausibly, and tags had been re-inserted.)
So, my speculation:
1. Apparently the pages in question are getting into your database at least once correctly, as they are able to display correctly once after that.
2. At some subsequent time, the page does not load into the editor correctly -- it's truncated. But even if not saved at that point, the database apparently now has a truncated copy of the page stored for that version.
So, apparently some process at the time of loading into the editor modifies the data in your database, or there's some other process running on the database that modifies the data in your database.
And it's somehow tripping up on something specific to do with tables. (It's possible that it's coincidental that it's tables that are affected for us -- perhaps those are the pages that, for us, are particularly long.)
Needless to say, this is highly alarming, and makes me doubt whether to edit any existing pages on the site, for fear I will lose a lot of work.
If you need more info on particular pages affected, I will be happy to provide it.
I am using Firefox 46.0.1 on Windows 7-64 current update. Though I don't think the client end matters, if not even the history has the last un-truncated good version of the page.
My reference Id: GWZ1031.
Graham
The page will look fine in read mode, but then invoke edit, and immediately the page is truncated at some point, with the rest of the page below that missing -- possibly a large amount.
Quitting the editor at that stage, ( without saving) results in display of the truncated page, the rest is missing.
It appears that data in your database is messed up at that point, because there's no recent version in history corresponding to the page in its final (non-truncated) state, which clearly there should be if that page once displayed properly.
Some more clues:
All of the three times, the page was truncated in a table (ie: table on the wiki page). Indeed, it was truncated in the middle of a row, with the result that that tr and table wasn't closed. So, in view mode, that resulted in the boilerplate rest of the page being messed up too, though that's merely a side effect of the not-properly-closed tags.
But it does indicate that this is an error that would normally not occur, even by user mistake, since I believe your editor save code goes to some lengths to make sure tags are closed. (I tried to manually delete closing tags in raw html mode, then save, and the page still displayed plausibly, and tags had been re-inserted.)
So, my speculation:
1. Apparently the pages in question are getting into your database at least once correctly, as they are able to display correctly once after that.
2. At some subsequent time, the page does not load into the editor correctly -- it's truncated. But even if not saved at that point, the database apparently now has a truncated copy of the page stored for that version.
So, apparently some process at the time of loading into the editor modifies the data in your database, or there's some other process running on the database that modifies the data in your database.
And it's somehow tripping up on something specific to do with tables. (It's possible that it's coincidental that it's tables that are affected for us -- perhaps those are the pages that, for us, are particularly long.)
Needless to say, this is highly alarming, and makes me doubt whether to edit any existing pages on the site, for fear I will lose a lot of work.
If you need more info on particular pages affected, I will be happy to provide it.
I am using Firefox 46.0.1 on Windows 7-64 current update. Though I don't think the client end matters, if not even the history has the last un-truncated good version of the page.
My reference Id: GWZ1031.
Graham