J
With the latest nightly version, it is possible to reposition the divider pane between the messages list and their contents: However, as soon as you restart the app, this gets reset to the default. It would be good to auto-remember the last used position instead.
J
Also, it would be good to thoroughly test !2201 with Test Center while you're at it!
J
It seems !2201 is supposed to fix this, in case some of you would like to thoroughly test it with Test Center!
J
It seems !2201 is supposed to fix this, in case some of you would like to thoroughly test it with Test Center!
J
Affected Version Version: Nightly flatpak only Distribution: Fedora 44 Also happens with development version: Yes Steps to reproduce Set Nautilus to single-click mode With the mouse, single-click a folder to enter it Do not move the mouse, and try to directly click the folder item that appeared under the mouse Expected Behavior Step 3 should activate the item under the mouse click. Actual Behavior The item requires two clicks: the first click selects the item, the 2nd click actually activates…
J
It seems to me like there might be a typo, presumably there's a missing space for %sallowing → %s allowing?
J
!2204 seems to provide great performance improvements in my case (testing with the "Test Center" app), I'd be curious to know if others are seeing similar speedups as well.
J
Presumably this is meant to solve #3374? If so, would probably be good to mention in the commit message. So far in my testing, the improvements are extremely promising. After flushing the kernel's disk caches: On my laptop, testing with an external USB SATA HDD loading a 1500 PNG screenshots folder: This branch loads the folder instantly Nautilus 50.4 would take 7.4 seconds to do the same thing On my desktop computer via RDP with its SATA HDD, if I ensure the process was not already running and…
J
Closing, as it seems in evolution-ews#339 you (or at least @fpaquet) are able to reproduce it within Evolution itself, so not a GNOME Calendar issue.
J
Isn't it possible to ask the scrolledwindow for the scrollbar widget (or get to the scrollbar widget somehow) and then gtk_widget_get_width on it, when the always show scrollbar setting is active?
J
For what it's worth, it seems at least GNOME OS will enforce this by default through @vixalien's gnome-build-meta!5046.
J
Is this all that was needed to resolve #896, or are there additional smart tricks planned for it?
J
play button was the square at the beginning, and then became the red sign for an error That symptom seems like a duplicate of #464 and #783? I think we really have to stop overloading that button with triple-duty. As I say in #783, leave error reporting to a separate widget, like a toast or regular system notification with a Show Details button.
J
Small detail, if it remains within the 1st commit (instead of a third commit preceding the current 1st commit): the 2nd paragraph of the 1st commit's message does not have consistent wrapping / line length. Otherwise, if it becomes its own commit again but preceding everything, then the problem will probably take care of itself.
J
Spinoff of #1187, but for ICS files instead of new events. Releated to #255 Feature summary Add an entry in Calendar's desktop file to import ICS files into Calendar. This will allow users to right click on Calendar's icon in GNOME Shell to open Calendar, then immediately open the file chooser portal to allow the user to pick ICS file(s) to import into Calendar. Relevant links, screenshots, screencasts etc.
J
I personally tend to disagree with that decision (I really don't see how having a desktop launcher action for this is hurting anybody) but will abide by Philipp's choice and close this as 3. Out of Scope (which I think is semantically more appropriate in this case, because there is no existing "behavior" so to speak ;)…
J
Looking at this again, another theoretically possible UX solution could be to group calendars by account (and use their "display name", as GNOME Settings also allows setting it for GOA sources nowadays; non-GOA Google account EDS sources will be tricky though, as per #1291) and provide section separators like what was implemented in #1196 (closed). That might be doable in the quick-add popover's widgetry (but it might get messy real quick, what do you think @TheEvilSkeleton?), but I don't know…
J
Alright so re-reading Milan's reply here: we can already e_source_set_enabled (as we do when the user decides to show/hide a calendar); we can set a calendar as the "default" with the 2 methods you mentioned (as we already automatically do in GNOME Calendar, using the most recently used calendar as the default for event creation/importing, since version 49+ as per #1135 (closed)); …but there is no API to delete the fallback "On This Computer/Personal" source, right? The API docs don't specify…
J
Affected version Distribution, including version: Fedora Silverblue 41.20241107.0 Papers cersion: 47.0 Package format and distributor: Flatpak from Flathub" Summary Text selection doesn't follow the document's structure, making selection hard. Description On documents with two side-by-side "paper" pages per PDF page, the text selection is performed as if they were a single page. In Firefox's PDF reader, it works as expected. This makes selecting text harder than it should be. Here is the…
J
I am not entirely sure that it is a Poppler issue, given that it works just fine in Okular, as I demonstrated in https://gitlab.freedesktop.org/poppler/poppler/-/work_items/1547#note_3735644
J
Affected Version Distribution and version: Ubuntu 26.04 Package format and distributor: deb from Debian Papers version: 50.2 (flatpak 51.0 behaves the same) Summary Selecting (OCR generated) text in a multi-column layout selects lines spanning different columns, but if I open the same document in Firefox, it works fine and I can select a paragraph in a column. I expect that from any PDF viewer. Description Please find attached a simple sample document with the mentioned problem. test.pdf
J
Does the issue occur with Okular too? If so, it's almost certainly a Poppler issue like #296.
J
It's been three years since someone last tested; Is anyone even seeing this at all in recent versions like 51.x? Personally I don't think I've experienced it in 50.x and newer, but then I'm running Fedora Workstation (i.e. RPMs + Flatpak), not Silverblue. Is it only something that occurs on Silverblue / rpm-ostree? If so, I think it should be part of the ticket title and have the rpm-ostree label too.
J
Alright, thinking about this some more, forget the wild "HTTP2" idea for now… my "Pareto principle" hunch is now that: 80%+ of the (perceived) performance win here would probably be to download icons in the same order as the view (or filtered/sorted model as shown in the view), so that they naturally populate the top row and whatever the user may scroll to next (without those extremely long "unexplained" pauses that occur between icons of the top rows, while it secretly is downloading a hundred…
J
Tested with ranged creation of a new event (in an empty area of the week view's timetable), and this branch does work without crashing, unlike main. There is one visual regression though: when click and dragging the empty range to create a new even, this branch clears the selection as soon as you release the left mouse button because the quick-add popover widget shows up, which looks weird compared to main where the range remains highlighted while the popover widget is shown (until the popover…
J
Could you fork this repo (don't forget to set up your SSH pubkey in your user settings first) and make a merge request out of your patch, referencing this issue as the thing being fixed in the commit message, so that it can be reviewed? Patch files will get lost/forgotten.
J
Maybe, but I kinda doubt it, because this one here has been happening forever even in version 50.0, whereas #1659 did not happen in 50.0 and only started happening with one of the commits mentioned over there…
J
Another set of symptoms with this bug is that you'll get really strange behavior when you drag another timed event just before vs during the affected event, as shown in this video (using the New York City timezone on my system, but your purple event is still on the Berlin timezone): Another thing that can be observed in that video is that the purple "Weekview day offset regression" sample event is faded out (semi-transparent) because the backend knows it is already in the past while I am…
J
I tried bisecting this, but unfortunately end up with imprecise results, because at least 5 of the commits in Georges' big GListModel refactoring MR at !742 (merged) do not compile, so I had to skip them: $ git bisect bad Bisecting: 4 revisions left to test after this (roughly 2 steps) [039b75755e23ead872f8c13cdebef3322ea97120] views/month: Port to GListModel $ git bisect skip Bisecting: 4 revisions left to test after this (roughly 2 steps) [c635db34957359f4857ab414781638a784548aa7] views/week:…
J
Actually one of the things I was waiting for was the fix for #989 (closed), which indirectly happened in 51. Now however, re-reading the comment I wrote in this MR's code here, I'm thinking it could make more sense to reconsider it only after @jcardullo's !808 lands, so that we can test whether the search range limit can be even more generous…