Replies: 49 comments 65 replies
|
What: Systematic deterministic (offline) tests for the storage providers that currently have unit coverage only. For each such provider, pin (a) the directory listing and stat parser against a captured real response, and (b) the HTTP-status-to-error mapping (401/403/404/409/429/5xx) with the correct retryable flag. Why: A change on a provider's side or in a third-party dependency (response shape, a new status code, a WAF quirk) should be caught by a failing test rather than by a user. This is already the pattern on the freshest fixes; the ask is to make it systematic across the long tail. Where: providers (backend), reflected in the Testing & Verification matrix. Legend: 🟢 pattern already applied · 🟡 pending (this request). Already applied as examples: WebDAV 404/409 upload mapping 🟢, OpenDrive size-limit error 🟢. Providers in scope (currently unit-only):
Context: came up in #322 while documenting test coverage. |
|
Done for v4.1.0: the deterministic offline test net now covers all 16 unit-only providers (one reviewable, behaviour-preserving commit each; Swift skipped, its only profile Blomp is inactive). Details and commits on the tracker: #361 (comment) The sibling idea (a single pre-release smoke entrypoint) stays open as a separate tracker item. |
|
What: A single pre-release smoke entrypoint. One command/script that runs the deterministic suite (Rust plus frontend) and the lab-backed integration tests, skips cleanly when credentials or Docker fixtures are absent, and prints a pass/skip matrix. Wire it into the release checklist as a pre-tag step. Why: Makes "run the tests before every release" a single reproducible action and makes coverage gaps visible at release time. It also lets a contributor run the deterministic core with one command without needing any lab credentials. Where: CLI / dev tooling, the release checklist, reflected in the Testing & Verification matrix. Legend: 🟢 exists · 🟡 pending (this request).
Context: came up in #322 alongside the published test matrix. |
|
🟢 Rolled forward to v4.1.0. v4.0.9 is out (release notes), so this thread now opens to the v4.1.0 window instead of starting fresh; the title and intro are updated to match. On what was raised here: the two suggestions in the thread (systematic provider determinism tests, and a single pre-release smoke entrypoint) did not land in v4.0.9 and are carried into the v4.1.0 tracker (#361) for triage, so nothing is lost. Everything that shipped in v4.0.9 is in the 4.0.9 tracker (#349). New enhancement ideas for v4.1.0 are welcome, one per comment. |
|
What: A Why: Creation is the one common action still missing from the interactive loop; surfacing it as Where: CLI ( 🟢 Planned for v4.1.0 (tracked in #361). originally from #311 (comment) (@EhudKirsh) |
|
What: Reorder the interactive Why: Where: CLI ( 🟢 Planned for v4.1.0 (tracked in #361). originally from #311 (comment) (@EhudKirsh) |
|
I noticed that |
|
While at it, please make |
|
Where: Dual panel when connecting OAuth profiles. What: Please add support for Crypt to OAuth profiles. Why: Right now, it's possible to set up a Crypt password in the Quick Connect page of OAuth profiles, but it doesn't actually show the |
|
Thanks, but in v4.1.1 the Crypt wrapper fails to |
|
I realise now that it's not to do with OAuth, but with the Crypt wrapper being On (blue) inside a folder with an unencrypted name. |
|
🟢 Shipped in v4.1.3. The crypt-overlay scope work landed: the overlay now shows plaintext outside its bound subtree and decrypts only inside it, across the GUI, CLI and MCP. That is exactly the "Crypt on inside a folder with an unencrypted name" case you isolated. Please retest on v4.1.3 and confirm on #390, which stays the tracking issue for it. Thanks for narrowing it from OAuth down to the overlay scope. |
|
Thanks, everything in this comment is fixed now. Changing directory works in both directions, both when Crypt is On and when it's Off, both inside and outside the Overlays Remote Path, thanks to it, and in OAuth as well. |
|
Thanks for confirming, glad it holds in both directions, with crypt on and off, inside and outside the Overlays Path, and under OAuth too. |
|
Where: AeroSync Compare. What: I uploaded a local folder to a cloud drive, and AeroFTP successfully encrypted it with Rclone Crypt (in #364). Now I want to be able to schedule regular automatic syncing for that folder with AeroSync. I tried the Canary Mode, and it told me that it will upload a byte size which happens to be that of the whole folder. Basically, none of the local files are detected at the destination. I went to AeroSync Compare and realised what happens. AeroSync Compare still doesn't take Crypt into consideration. So please make it so that it does. Why: To be able to schedule syncing with AeroSync while using Crypt without re-uploading everything each time, just update the changes. |
|
Captured and tracked on the 4.1.4 tracker (#408). For the record on what already landed: v4.1.1 shipped the overlay-consistency fix (one crypt decorator at every resolver chokepoint, fail-closed) that came out of an AeroSync compare mismatch on a crypt profile. The remaining piece is the efficiency you describe: decrypt remote names and sizes at compare time so an incremental run transfers only the changed files instead of re-uploading the whole folder. The scope foundation it needs landed in v4.1.3, so sizing the exact compare entrypoint is next. Same goal as rclone sync over a crypt remote. |
|
I checked in v4.1.4 and this doesn't seem solved. AeroSync Compare still shows that the byte size difference is basically the entire folder when comparing a remote encrypted folder to its local source of truth (💻➜☁). |
|
Fixed on |
|
In 4.1.6, I'm getting this error for both Rclone Crypt and AeroCrypt: I get a similar, yet different problem with Filen, where AeroSync Compare measures the entire folder size as being out of sync. |
Fixed for v4.1.7, and thank you for reporting it as one item: the two halves turned out to be different root causes with the same consequence, and the crypt half was actually three. The short version is that Compare was lying rather than failing, which is the worst way for it to break, so an empty diff you read as "in sync" was really "nothing was compared". Your Filen observation was the decisive clue for the second half. Every upload stamped the transfer time as the file's modification time, and since that is the only timestamp a Filen account holds, the folder compared as entirely out of sync against the local folder it came from, permanently. That is also exactly why you saw it over WebDAV as well: the desktop bridge serves the same stored value. Details: #458 (comment) |
|
Where: AeroAgent > AI Settings > Models > '➕ Add Custom Model'. Why: This is yet another window that's still dark during a light theme. For consistency, it's best to have it follow the selected colour theme. What: Please make sure that this window is in light theme, as was the case in #270 (comment) with the Cryptomator window. |
|
Fixed for v4.1.1: the model add/edit window (and the "Available Models" list) in AI Settings now follow the selected theme instead of staying dark, matching the Cryptomator window. |
|
Where: The preview of text files in AeroFile. What: Please follow the theme here. Why: Consistency. |
|
Fixed for v4.1.1: the text preview now follows the Light and Ice themes instead of always rendering dark (toolbar, gutter, code and markdown all follow). |
|
Where: The preview of text files in AeroFile (same as last comment). What: Please increase the width of the scrollbar, as it's too small now. Alternatively, it'd be nice if I could add a custom ::-webkit-scrollbar{width:20px !important}Why: Maximise customisation, or at least increase the width of the scrollbar so it can be seen and dragged more easily while not taking any horizontal space that was used by text anyway. |
|
The preview scrollbar is widened in v4.1.1 (6px to 14px) so it is easy to grab, scoped to the preview so it does not take space elsewhere. The user-supplied custom-CSS-snippet idea (Obsidian style) is a nice one and is noted separately for consideration. |
|
Where: The preview of text files in AeroFile (same as the last 2 comments). What: Clicking on a line number should select all the text on its line. Why: It's intuitive and doesn't take any extra space. |
|
Captured for triage: clicking a line number to select the line, drag to extend, IDE-gutter style. Not in v4.1.1 yet, but it is on the list. |
|
I noticed that it's perfect in the <> Editor. In Preview it still struggles to drag to select lines in both directions, especially when going down. Also, unlike the <> Editor, Preview doesn't copy when pressing Ctrl+C, but only when right-clicking and clicking to copy. |
|
Thanks for the retest, good to hear the gutter line-select is solid in the |
|
Please update the Google Drive and OneDrive icons wherever they are (e.g. https://github.com/axpdev-lab/aeroftp#integrations) to their new official icons. Sources:
|
|
Good catch, both marks were redrawn in 2026. Refreshing our Google Drive and OneDrive logos (and the README integrations row) to the current official icons is captured on the 4.1.4 tracker (#408). Thanks for the sources. |
|
Please update the old Google Drive logo in the 'Connect to Google Drive' and ‘Sign in with Google Drive’ buttons in Google Drive’s Quick Connect page. |
Done for v4.1.7, and thank you for coming back on it, because the mark was stale in more places than the two buttons. The 2026 refresh you prompted updated the canonical logos and the grid icons, but three components each carried their own private copy of the old marks and were invisible to that change: the Quick Connect page you point at, the Add Service protocol list, and the session tabs. The first two were on screen; the third was already unreachable, so it was stale code sitting there waiting to be brought back by a later edit. All three now take the mark from the one place it is defined, so this class of drift cannot happen again. One visible consequence worth flagging before you see it: the Google Drive buttons are no longer red. The mark you asked us to retire was a flat six-colour form, which is why it could be shown as a white silhouette on a coloured button. The 2026 mark is three gradients, and Google's brand guidelines do not allow restating it in a single flat colour. Google publishes a neutral button surface for exactly this case, so those two buttons now use it and carry the mark in full colour. Every other provider keeps its brand-coloured button, since each of those marks is a single flat shape that a white version does not distort. Details: #458 (comment) |
|
Where: The Add Service and Quick Connect pages of pCloud Drive and AWS, among likely other places. What: Please rename 'pCloud' to 'pCloud Drive' and 'Amazon S3' to 'Amazon Web Services (AWS)'. Why:
|
|
All three are fair: pCloud is the company and the app is pCloud Drive, 'Amazon Web Services (AWS)' reads truer than 'Amazon S3' with the S3 service carried by the badge, and 'aws' should match in the Add Service search, which returns nothing today. Captured on the 4.1.4 tracker (#408). Thanks. |
|
Where: AeroSync Compare What: Please add a measure in %, as in, what percentage of the byte size or the number of items which are out of sync. Why: This will make it easier to immediately understand the extent to which folders are out of sync with each other. |
|
Done on |
|
Where: The '🗑 View Trash' of OpenDrive API, kDrive and probably others. What: I see a red '🗑 Permanent Delete' button and an orange-brown '⚠ Empty Trash' button. Why: Simplify and clarify. |
|
Done for v4.1.6. They are two genuinely different actions, so rather than merge them we added tooltips that spell out the difference: Permanent Delete removes only the selected item(s) from Trash, Empty Trash purges everything. Applied to the six dual-action trash managers (OpenDrive, kDrive, pCloud, Yandex, Nextcloud, and Dropbox, whose Permanent Delete / Empty Trash buttons also landed for v4.1.6). #422 (comment) |
|
Where: AeroSync. What: Please make sure that the custom settings in AeroSync do not get deleted when switching between the Plan and Sync tabs. I'm not sure how I'm supposed to be able to schedule a Plan with Update and other settings if they're deleted when switching to the Sync tab. Why: To be able to schedule with AeroSync with particular Plan settings. |
You were not supposed to, and nothing was deleting them on purpose. The dialog renders |
|
Where: AeroSync Compare. What: When scanning at the start, please show a timer of how long the comparison has taken so far, and how many files, folders and total items were scanned and their total byte size. Why: These are all stats that will be summarised later anyway, so they might as well become a progress bar while the user waits for the comparison to finish. |
Done, and it turned out most of it was already on the wire: every compare walker has emitted a |
|
Where: '🗑 View Trash'. What: Be able to sort ascendingly and descendingly by name, byte size and time & date deleted. Please also add a Type column to sort between files 📄 and folders 📁, as I see their icons there, so clearly AeroFTP already supports that ability. Why: To be informative and quickly find the items that a user is looking for. |
Done, for every provider at once. '🗑 View Trash' was not one dialog: seventeen provider components each carried their own copy of the same table, and not one of them had any sorting at all, so rather than patch seventeen tables this is now one shared table that all seventeen render. Type, Name, Size and Deleted all sort both ways. Folders and files only split apart under Type; a folder has no size of its own, so it sorts last under Size whichever way the column points, rows with an unreadable date land last under Deleted, and every tie falls through to the name so the order cannot shuffle between renders. Sorting by file format specifically is not in: the trash listings do not all report a type beyond file-or-folder, so that column would be guessed from the extension rather than known, which is a different feature. On |
|
Where: '🗑 View Trash', same as the last comment. What: Please move 'Name' to the left so it's either centred on its column or at least on the left side of it. Why: It's like the tables in the Add Service and My Servers pages, which can align the headers of their columns. |
Done, by moving what was in the way rather than by nudging the label. The file/folder icon used to live inside the Name cell, so the names began well to the right of their own header while the header itself sat hard against the checkbox column. The icon now has its own Type column, the one in the previous reply, and the Name header lines up with the names underneath it. On |
|
Where: '🗑 View Trash', same as the last 2 comments. What: Please show the timestamps in a single row by widening its column and narrowing the Name column. Why: To see more rows of items on the screen at the same time. |
Done, exactly that trade: the Deleted column is wide enough for a full date and time and no longer wraps, and the width comes out of the Name column, which has it to spare. Same shared table as the other three, so it is true in all seventeen trash dialogs rather than in the one you were looking at. On |
|
Where: '🗑 View Trash', same as the last 3 comments. What: Add rubber-band box support to '🗑 View Trash'. Also support using the Ctrl key, like in file managers, to add selection of more items while keeping existing items selected, allowing selecting items that aren't next to each other by using the rubber-band box multiple times. Why: To make it faster to select items. |
Done. The rubber band reuses the marquee code written for AeroFile in #270, whose own documentation gave the trash view as its example of somewhere the behaviour is deliberately switched off. Ctrl and a plain click both add without clearing, since these rows are checkboxes, and Shift extends a range from the last row you clicked. The anchor stays put after a Shift-click, so a second one re-extends from the same origin instead of walking away a click at a time, which is what makes picking several ranges in turn work, the part you asked for. One limit worth stating rather than letting you find it: dragging past the bottom edge selects what is visible and stops there, because the scrolling element belongs to each provider's dialog and the marquee's auto-scroll reads the wrong one. Scroll first, or hold Ctrl and drag again to add. On |
|
Where: My Servers and Add Service pages. What: Please add a '🔒 128-bit' or '🔒 256-bit' badge for all the protocols of MEGA and Filen if they all support client-side encryption. Why: To be informative. Any protocol and profile that has client side encryption should clearly indicate this. |
Half of this existed and half was a real gap. Add Service has carried an "E2E 128-bit" / "E2E 256-bit" chip on MEGA, Filen and Internxt for a long time; My Servers said the same thing only in prose, in the subtitle under the profile name, which is not where the eye goes. It now carries the badge next to the name, and both surfaces read one list with a test holding them to it, so they cannot drift apart again the way they were free to. One correction to the premise, and it is the interesting part: not all of MEGA's protocols encrypt client-side. MEGA S4, their paid S3-compatible object storage, is a profile with |
|
Where: Find Duplicates in AeroFile. What: Please show a progress bar counting how many files, folders, byte size, depth and time were scanned, similar to #347 (comment). Why: Pretty much anywhere that the user waits for a while, it’s important to show the progress so they can guess how long they have to wait, and also to confirm that there’s no infinite loop bug. |
Done: files, folders, bytes and depth during the walk, then a determinate bar during the analysis pass, with a clock throughout, throttled to 120 ms so the reporting cannot slow the scan it reports on. Non-identical mode used to let the dedupe engine do its own separate walk; it now takes the file list from the same walk as exact mode and reports its signature pass back through a callback, so both modes are visible and cannot drift apart. On the group counter: it is the one already in the summary bar. During the scan there is no honest number to show, because a group does not exist until the signatures that form it have been compared, and a running count would only be telling you how far the comparison had got, which the bar already says. On |
|
Where: Find Duplicates in AeroFile, same as the last comment. What: Please add Copy buttons next to the folder paths and names of images. Possibly also buttons to open these in their default apps. Why: To not have to rely on an OCR app, which is inaccurate and takes too long to be able to find these duplicates externally to examine which I’d like to keep and which to delete. |
Done: a copy button on the file name and another on the folder path of every row, plus one to open the file in its default app and one to reveal it in the file manager. The |
|
Where: Find Duplicates in AeroFile, same as the last 2 comments. What: Please add a column that specifies the hash of each file, or maybe write it under each file, and add a global checkbox or switch that toggles showing and hiding the hashes. Also, please enable sorting groups of duplicates by how similar their fuzzy hashes are. Why: I tried the new feature of finding non-identical duplicates in AeroFile. Thank you for adding it, by the way. |
All three, and the middle one found a bug that made it impossible as asked. The number the dialog printed as "dist" was On top of that: every group now carries its members' own signatures, pHash and SimHash as hex and TLSH as its own string, behind a "Show hashes" toggle, with the byte-identical BLAKE3 shown in exact mode. The Sort control offers Similarity, which puts the identical groups first and then the fuzzy ones closest-first. And there is a Fuzzy cutoff field, blank meaning the per-type defaults. That last one had never reached the engine: the command took no threshold and always passed none, so those defaults, images 10, text 3, other 100, were the only thing anyone could ever run. Which means your sweet-spot question is now something you can answer by hand rather than a thing to wonder about, and the case you described, the same image with text on it, is exactly the one that should separate at a tighter cutoff while a re-encode survives it. I would be interested in what you find. On |
|
I just found something interesting already in v4.1.6: two images that look exactly the same to the naked eye both had roughtly the same byte size to the kB, but different cryptographic hashes. Find Duplicates found them. Another pair of images looks very similar, but is clearly different. The smiles are different, and there’s text showing in one but not the other. |
|
Where: Find Duplicates in AeroFile, same as the last 3 comments. What: Please write the Why: It shows that AeroFile maximises efficiency by using the fastest reliable hashing method that's available, and not bothering to calculate full 256-bit checksums, since it's not hashing a password that needs 256 bits of entropy, which offers 128 bits of security against Quantum, since 64 bits is presumably too little against Quantum, but 128 bits is sufficient for collision resistance, just like with AeroChunk IDs. By the way, this reasoning doesn't translate to fuzzy hashes, because they need to remain long hashes to compare fine differences between each other. So for fuzzy hashes, length offers greater resolution to measure similarity and distance. |
|
Where: Find Duplicates in AeroFile, same as the last 4 comments. What: Be able to tick which duplicate files to delete and which to keep. Right now, it chooses which file to keep and which to delete, and doesn't let ticking one of the duplicate files to delete. For example, I tried this with a group of 3 duplicates, and 1 of which I couldn't tick to delete. There's even a 'Select All Duplicates' button that doesn't actually do as it says and considers one file as an original so it's not ticked. Maybe it should be renamed to 'Tick all duplicates except one from each group' or something like it, and a 'Select All Duplicates' button should be added to do as it says. Why: Sometimes it makes sense to delete a duplicate from a specific folder (they're not all in the same directory) or delete the one with |
|
Where: Find Duplicates in AeroFile, same as the last 5 comments. What: Please add a tiny preview square or rectangle to the side, right or left, of each group of duplicates. It would preview images like AeroFile does when viewing ‘Large Icons’. Clicking on a preview should expand the view of the image or text file like in AeroFile. Why: Be able to clearly see which file is the duplicate, be able to view or read it, and even edit it in AeroImage or the <> Editor of AeroTools. |
|
Where: Find Duplicates in AeroFile, same as the last 6 comments. What: Please add a scrollbar there on the side. At least 14px in width ideally, maybe even wider. Why: There are plenty of rows to scroll through. |
|
Where: AeroFile 📁 in Icons and Large Icons view. What: Please don't hide the edges of images. Each icon should show its image in full. Why: I like being able to see the edges, and that's also what File Explorer already shows. |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Welcome to the AeroFTP wishlist. This is a single permanent thread that collects enhancement suggestions from the community and triages them into upcoming releases. It is not tied to one version: it rolls forward release after release, so there is no need to open a new wishlist each cycle. Just post here.
Suggestions raised here feed directly into the current open release tracker, where accepted items are scheduled and shipped work is logged. The tracker always points at the version being worked on right now.
🟢 The current open tracker is #370 (4.1.1). When a release ships, this thread stays exactly where it is and simply starts feeding the next tracker.
How to contribute
Please post one suggestion per comment so each one can be reacted to and discussed independently. A good suggestion includes:
Reactions on individual comments are welcome and count as "I want this too" signals. Free-form replies discussing trade-offs are also welcome under the same comment.
Scope
In scope:
Out of scope:
Triage
For each release I will:
Reactions on comments inform priority. Tie-breaks favor items that improve daily-driver ergonomics for tester profiles (existing community testers, documentation drilldown, error message clarity).
Items already shipped are considered done and do not need to be re-raised. If an earlier suggestion was acknowledged but never shipped and is still open, please post it again as a fresh comment with a one-line
originally from <link>trailer, so reactions stay meaningful for the current cycle.Thanks in advance for the input.
All reactions