David Cook
David Cook
The pdfrw errors are happening because the tokenizer isn't handling inline images, and when the intervening binary data includes a left angle bracket, it tries and fails to parse it...
See pmaupin/pdfrw#67
Turns out all of these will have to be fixed in pdf-redactor. Reference links: http://stackoverflow.com/questions/39927639/is-it-possible-to-use-libtiff-to-decode-ccitt-encoded-data-when-the-length-is-no https://issues.apache.org/jira/browse/PDFBOX-2163
Panic-free Drop implementations might be a good intermediate goal, as a second panic during unwinding will abort
This is an exciting avenue! One thing I like is that it could also catch some of the issues the failpoints and crash recovery tests do, but reproducibly, by chopping...
I had some stray thoughts on this I wanted to jot down. The ALICE paper says they "serialize system calls in the order they were issued", with the assumption that...
Hmm, perhaps we should try fsyncing the directories and configuration file when creating a new database. I'd recommend turning on debugging symbols in your release profile, so that the static...
I added fsync calls on directories, and I still got the same error messages, so that wasn't it (see below). FWIW, I had to patch the ALICE tool in a...
I re-ran it with the fsync changes, and got a similar report about durability of operations 0 through the final write to "db". I think the issue is that the...
Double-recovery bugs have been hard to root out in the past, it would be good if we could apply ALICE to the second crash/recovery, as this would crank through a...