unfortunately it simply was not reliable enough
and maintaining it is simply adding more complexity to the codes
moreover it was only possible to passthrough regular non-chunked uploads
maintaining it is an unnecessary complexity
it's a feature that doesn't scale too well anyways
also renamed "queryDbForFileCollisions" to
"queryDatabaseForIdentifierMatch"
and updated config description accordingly
this should also now properly free the internal onHold Set
instead of converting from the default latin1 using Buffer
changing busboy config was not possible with express + multer,
so i did not notice it's instead possible with hyper-express
unfortunately to capture multipart errors, we have to wrap the entire
callback function with try-catch block
but it appears overall processing is still slightly faster than
expressjs + multer
this should have better lifecycle and use less memory over time, since
we can define max items in cache
at the moment hard-coded to 10 cached pages (inclusive of nojs version
if ever generated)
we used to store number directly into the string size field, and
during the conversion it seemed to always add ".0" at the final string,
probably because the driver or sqlite3 itself assumes float
please run yarn migrate after pulling this commit
if you skip converting the DB, file duplicates check will fail to
function
and in the future im planning to do size statistics in bigint, which
will also fail if not converted
this will display all information recorded from the specified file, but
only to the users that own them (it requires token)
this page also has a delete file button, allowing us to provide link to
this page for sharex deletion url option
once again, this is only for authenticated users, and will only show
file that the users own, unless said user is a moderator or higher
db is now initiated via utilsController, and gets re-used by main
lolisafe.js script and the other controllers
this should now make sure we only have 1 active db connection pool ever
this is a pretty big refactor due to how we used to reference db,
but it should functionally be identical
this should properly dispose of unfinished write & hasher streams on
any errors if they have been initiated
also do content-length header check a bit more early
