The service will now automatically clean up any leftover data from any
inactive & unfinished chunked upload attempts.
Updated config.sample.js for updated description and new sub-option
for chunkSize option.
UPDATE YOUR CONFIG FILE IF YOU USE CHUNKED UPLOADS!
Read more about this at the end.
Added new dependency: blake3
Hashes will be created as the uploads are being written to disk.
With exception for chunked uploads!
For them specifically, their hashes will be created as they're being
rebuilt into a single file.
Should still be a lot better than the previous case where it had to
re-read the already written files.
To support that feature, added a new file
controllers/multerStorageController.js.
It's just a custom storage engine for Multer.
chunkSize option now allows setting max chunk size from config file.
Previously it was hardcoded to 95MB, but assuming you have paid
Cloudflare plans, you can actually have up to 500MB.
Also moved the option to be after maxSize and before urlMaxSize.
Made a lot more sense to me this way, as chunked uploads only work on
regular uploads.
Updated v1 version string and rebuilt client assets.
This adds Cookie Consent library at public/libs/cookieconsent.
Added views/cookiepolicy.njk.
Added a new config option cookiePolicy. Disabled by default.
Updated views/faq.njk to use variables wherever applicable.
The variables are initiated at views/_globals.njk.
The said file also contains variables used at views/cookiepolicy.njk.
Restored messages about Cloudflare to FAQ.
They will be shown only if config.cloudflare.purgeCache is enabled.
Statistics will now capitalize first letters of the keys instead.
Updated background color of SweetAlert modals.
Moved table styling from dashboard.css to style.css (global),
since table will also be used in Cookie Policy page.
Bumped v1 version string and rebuilt client assets.
Restore disk usage stats even when config.linuxDiskStats is off
The said config will now only toggle the 'extended' stats which are
disk usage of each directories within the uploads directory
The option is disabled by default.
Meaning you need to explicitly set it to true to restore
the old behavior.
The one that uses both "df" and "du" binaries.
I found them to be very slow with 100k+ uploads
on my ancient potato server.
... and optionally videos using ffmpeg (still experimental).
Users can choose whether to strip tags of their uploads or not from
the home uploader's Config tab (safe.fiery.me will have it disabled
by default).
The behavior will also be applied to the downloadable ShareX config.
Server owners can choose to force either behavior.
Make sure to add the new config from config.sample.js.
---
Fixed all instances of "e.i." to "e.g.".
My English sucks okay.
Bumped v1 version string.
Updated sample API rate limits.
This will pretty much be the same ones used live in safe.fiery.me.
This rate limits ALL API calls to 10 requests per second,
but apply stricter limits to login & register endpoints, which are
2 requests per 5 seconds.
Also apply stricter limit to album ZIP download endpoint to
4 requests in 30 seconds.
Also removed forcing 200 HTTP status code from the error responses,
cause front-end will now handle any HTTP status codes properly.
It was previously set to 200 cause frontend couldn't handler
errors properly.
On a side note, rate limiting all API calls is important due to the fact
that any token-based endpoints can be used for brute-forcing tokens.
Some server firewalls can also be used to ban possible brute force
attacks through actively monitoring the HTTP server's access logs,
so you may also want to consider that kind of solution for your site
instead.
Description in album public pages will no longer use h2 tag.
Descriptions that can go up to 4000 chars kinda made no sense to use
that tag.
Use Nunjucks' built-in nl2br tag to replace newlines with <br> tag.
Removed unused macro from faq.njk.
Better extensions filter list in URL uploads tab.
Updated config.sample.js with sample usage of the URL uploads extensions
filter.
As the title says, this commit is a massive overhaul.
I've rewritten/restrucuted almost everything in the controller scripts.
Because of that, there's a considerable possibility that I've broken
something somewhere.
Notable changes:
Added temporary uploads.
Removed file name length changer from dashboard,
in favor of an equivalent in homepage config tab.
This allows non-registered users to also set file name length.
A bunch of other undocmented stuff.
I don't know, I'm too tired to remember them all.
* Updated path resolving for upload folder option.
This SHOULD now allow using an absolute path, even those outside of
the lolisafe installation.
* Added a config option at uploads -> generateThumbs, named placeholder.
It's a string option that lets you set path of the placeholder image
for files whose thumbnail could not be generated.
It defaults to the old hard-coded path when set to falsy value.
* Fixed thumbs script not using upload folder from config file.
* Added filters for thumb generation error handling.
This is used to ignore some common error messages,
such as file formats not being supported.
tokenController.js + authController.js:
+ Added a standalone function to generate unique token.
Despite tokens being 64 characters long, meaning the chance to generate
the same token twice have very small chances, I would rather not leave
it to chances.
+ Some spacings.
config.sample.js:
+ Self-explanatory.
config.sample.js + uploadController.js:
+ Added option uploads > storeIP to toggle whether to store uploader's
IPs into the database.
uploadController.js + dashboard.js:
+ Added IP column when listing all uploads.
+ Improved album query when listing uploads. In addition, no longer
query album when listing all uploads.
+ Delegate some tasks to client when listing uploads to save server's
processing power, kek.
Such as building the file's full URLs, and assigning album/user names.
_globals.njk:
+ Bumped v1 version string.
utilsController.js:
+ /api/stats: Cache invalidation requests will now only store timestamp
of the request instead of purging the cache entirely.
Subsequent requests to the API will no longer attempt to generate stats
if a previous request before it is still generating, since it can
take a while in big sites.
So instead, the cached stats will be returned to them, regardless of
whether it's still valid.
This should avoid "race condition" in sites with multiple admins.
* dashboard.js:
+ Added an appendix into "Delete by names" feature in dashboard to
inform staffs that they can use the feature to delete files by other
users as well.
+ Added comments explaning the need of page.domClick() function.
+ Addition to the change regarding /api/stats route: If there is no
cached stats, but a previous request is still generating, inform
the user about it.
config.sample.js + lolisafe.js:
+ We now support configuring rate limits through config file.
No more hard-coded rate limits.
Don't forget to update your config.js accordingly.
package.json + yarn.lock:
+ Updated dependencies.
Updated fontello (added block and doc-inv icons).
Upload results will now show either doc or block icon on top of the
file name depending on the status of the upload
(unless the uploads are images, in which case they will still show
thumbnails instead).
Added support for customizable timeout and chunkSize options for ClamAV
scanning to the config file.
Bumped v1 and v3 version strings.
Yes.
This gets rid of HEAD request prior to downloading the URL.
We will no longer check for Content-Length header, instead we will
forcibly limit maximum download size for the download stream to the
configured value.
So assuming someone try to download a bigger file, it will still try to
download up to the configured size, but then fail.
This will also speed up the general download process since sending HEAD
request delayed the whole operation.
Added extended support for URL uploads.
Namely URL proxy support and separate extensions filter (as in separate
from the primary extensions filter).
There's also a new option to set a disclaimer message that will be
printed underneath the URL uploads form.
Trust proxy is now toggleable from the configuration file.
I think they should only be enabled when you're behind proxy such as
Cloudflare or Incapsula.
I'm not sure how it behaves with only a bare nginx reverse proxy though.
Empty files can now be filtered.
Sorted preset extensions filter in config.sample.js.
Rephrased some options in config.sample.js as well.
maxTries now default to 3 in config.sample.js.
Various other small changes.
Added a new config entry named "cacheFileIdentifiers". More info in config.sample.js file.
Improved some descriptions in config.sample.js file.
Added some CSS animations wherever applicable.
Bumped v1 version string.
Reworked unique name generator to prevent the same unique identifier from being used if it was already used with a different extension (e.i. If a file named aBcD.jpg already exists, then files such as aBcD.png or aBcD.txt may not exist).
This is mainly to deal with the fact that thumbnails are only being saved as PNG, so if the same unique name is being used by multiple image/video extensions, then only one of them will have the proper thumbnail.
If you already have existing files with matching unique name but varying extensions, unfortunately you can only deal with them manually for now (either allocating new unique names or deleting them altogether).
Added a new config option to filter files with no extension.
Files with no extensions will no longer have their original name appended to the allocated random name (e.i. A file named "textfile" used to become something like "aBcDtextfile", where "aBcD" was the allocated random name. Now it will only just become "aBcD").
In relation to that, utils.extname() function will now always return blank string if the file name does not seem to have any extension.
Though files such as '.DS_Store' (basically anything that starts with a dot) will still be accepted.
Examples:
.hiddenfile => .hiddenfile
.hiddenfile.sh => .sh
.hiddenfile.001 => .hiddenfile.001
.hiddenfile.sh.001 => .sh.001
Simplified error messages of /api/upload/finishchunks.
Most, if not all, of the error responses for /api/upload* will now have HTTP status code 400 (bad request) instead of 200 (ok).
I plan to generalize this for the other API routes in the future.
Updated home.js to properly handle formatted error message when the response's status code is not 200 (ok).
Bumped v1 version string (due to home.js).
* Hide notice about max size if it's equal to max size of regular uploads.
* Show notice about DuckDuckGo's proxy if it's turned on.
* Also updated description in config.sample.js to further emphasize about the limitation of using DuckDuckGo's proxy.
* When gitHash in config.js is set to true, latest commit hash of the currently enabled git repo/branch will be displayed in home and nojs uploader pages.
* Error pages can now be configured with errorPages option (their root directory and their file names).
* Added an experimental virus scanning feature using ClamAV. This has only been tested with an Ubuntu machine.
* File extensions will now be parsed with path-complete-extname module. This will ensure extensions such as .tar.gz are properly parsed.
Notice: It may take a minute or so to start the safe with virus scanning, as apparently the module takes a while to create the engine. I'm guessing since it'll be loaded to memory? Either way, once the engine is created, everything should work fine. Virus scanning should also not have that much of an impact to the upload time.
Summary:
* Added 3 new sites to README.md.
* Added some new file extensions to filter.
* Some additions to nginx sample conf files.
Changes such as delete buttons and whatnots for the dashboard were not added.
Added DuckDuckGo's proxy support for "Upload by URLs". Make sure you add the new option in config.sample.js into your config.js.
This may be considered a hack and not supported by DuckDuckGo, so USE AT YOUR OWN RISK.
Credits to Proxy#1337.
* Added upload by URLs. It has its own max size (config.uploads.urlMaxSize), make sure your config matches config.sample.js.
Here's a brief video showing it in action: https://i.fiery.me/CUhQ.mp4.
* /api/upload now supports uploading by URLs. Devs will only need to POST a JSON request containing a key named "urls", which is an array of the urls to upload.
* Added file extension filter to /api/upload/finishchunks.
* Added proper total chunks size check to /api/upload/finishchunks.
* Various code improvements.
* Added Cloudflare purge cache support. Check configuration sample at config.sample.js.
When it's enabled, whenever files are being deleted, it will send a POST request to Cloudflare's API to purge cache of the deleted files.
This adds a new dependency called "snekfetch". It's lightweight though.
* uploadsController.delete() will now wrap uploadsController.bulkDelete() instead.
NOTICE: Please update your config.js. Use config.sample.js as the template.
There were a couple of renames and restructures.
* Album zipper API route will now internally save its state when it's generating zip files, and any subsequent requests will silently be "postponed" until the first spawned task is finished. This will guarantee that there are no multiple zipping tasks for the same album. The method may seem a bit hackish though.
* All instances of console.log(error) were replaced with console.error(error). This will guarantee that any error goes to stderr instead of stdout.
* Deleting file by names will now properly remove successful files from the textarea. There was a logic flaw.
* Failure to generate thumbnails will no longer print the full stack, but instead only the error message. It will also then symlink a template image from /public/images/unavailable.png (it's only a simple image that says that it failed to generate thumbnail).
This haven't been tested in Windows machines, but it'll probably work fine.
I thought of adding a new column to files table which will store information whether the thumbnail generation is sucessful or not, but oh well, I'll go with this method for now.
