We have a google product feed which we've just moved to xml on a scheduled upload (until now I have been taking one of the affiliate feeds and manually tweaking it and uploading as a .txt)
All 12k+ products upload successfully. *some* of them however display the description (and only the description) as #########
No error is reported in the upload log google provide, the only errors that throws back are that we sometimes don't supply a delivery time etc.
We already remove some special characters (carriage returns, new para, etc) and we can't find a single character which only appears in the ####ing failures. We do have a huge variety of the *"-!(%x/ things but none of them seem to be the issue.
We've stripped all html so its also unlikely to be something copy and pasted into the text.
We're about to try truncating descriptions to <1000 characters (although I'm almost certain its not this and Google state 10k characters for the description).
This has happened to me occasionally with the manual uploads but a reupload has always fixed it. I've forced a refresh of the feed twice now and the problems still there, after more than 24 hours, so it shouldn't be a delay in showing the new data.
After this my next step will have to be a rewrite in the cms of any descriptions causing issues, but thats going to be hard to find them, a pain to do, and even harder to ensure we don't have the same problem again, so its not ideal.
There must be something really simple I've overlooked. Anyone any ideas what?
<added> - I'm also going to check whether our " might use two different encodings for some reason.
