* Remove `banned_ip_for_download?` config option. This isn't something that usually needs
to be configured.
* Replace the `ipaddress` gem with `ipaddress_2`. The `ipaddress` gem has several methods
we need (`link_local?`, etc) that are only available in master because the gem hasn't had
an official release in several years. `ipaddress_2` is a fork that is more actively
maintained.
This was a hack to deal with the Cloudflare check sometimes being slow
or timing out during tests. The call to https://api.cloudflare.com/client/v4/ips
could hang if there were IPv6 connectivity problems. If this happens, make
sure that IPv6 is configured properly and that `curl -v --http1.1 -6 https://api.cloudflare.com/client/v4/ips`
works.
Fixes the banned IP check not being applied when sending the HEAD
request for is_cloudflare?.
Also fixes the `#size` method not using the uncached url (which meant
the bookmarklet could report the wrong filesize on artstation uploads).
Bug: if an upload timed out while downloading the file, Upload#process!
would catch the error and attempt to retry, but since the upload was
already in the 'processing' state, on the second try `process!` would
bail out immediately and leave the upload stuck in the 'processing' state.
Fix: remove the retry logic from Upload#process!. Let Downloads::File#download!
(which had its own retry logic) handle it instead.
Refactors the upload process to pass around temp files, rather than
passing around file paths and directly writing output to the local
filesystem. This way we can pass the storage manager the preview /
sample / original temp files, so it can deal with storage itself.
* Change Download::File#download! to return a temp file.
* Change DanbooruImageResizer and PixivUgoiraConverter to accept/return
temp files instead of file paths.
* Change Upload#generate_resizes to return temp files for previews and samples.
* Change Upload#generate_resizes to generate ugoira .webm samples
synchronously instead of asynchronously.
* Change `http_get_streaming` to write the output file directly,
instead of taking a callback.
* Track the filesize as the download progresses and abort when it
exceeds the limit.
* Don't save the Content-Type (it's not used anywhere).
Fixes an exception when attempting to upload a Pixiv image from a
deleted work. The download strategy tries to fetch the source data in
the course of rewriting the URL, which fails if the work has been
deleted from Pixiv.
Raise a BadIDError and leave the URL as-is (don't rewrite it).
Due to missing || operators, for Pawoo/Tumblr/Artstation uploads the
source wasn't changed from the direct image URL to the html page URL
after upload.
