https://bugs.kde.org/show_bug.cgi?id=422306
--- Comment #10 from MountainX <davestechs...@gmail.com> --- I was able to do further testing. I found some interesting things. 1. This issue is affected greatly by ~/.local/share/user-places.xbel. Deleting that file causes Dolphin to create a new one that contains only local storage devices and remote mounts (such as NFS) that are listed in /etc/fstab. This seems to be the minimal user-places.xbel that is possible in Dolphin. Comparing the minimal user-places.xbel to a "large" version with 30 user bookmarks, half of which have a customized icon, I see a difference in Dolphin startup time of over 600%. With the "large" user-places.xbel my Dolphin startup time is 2-3 seconds. KDE file dialogs are also similarly slow to come up. The the minimal version of user-places.xbel, KDE file dialogs open without any apparent delay and Dolphin has a minimal delay (I estimate half a second). The approximate half second delay may be due to the fact that my minimal version includes 10 NFS mounts. I create intermediate versions, each with varying amounts of my own bookmarks in Places (which increases the items in user-places.xbel compared to the minimal). With each increase, there is a corresponding longer startup time for Dolphin and a corresponding longer time for KDE file dialogs to come up. I had seen this behavior before, but now we have verified it in a more rigorous way. I am highly confident that user-places.xbel can have a strong negative impact on Dolphin start up times (and KDE file dialogs). The relationship is 100% causal and 100% reproducible. 2. I believe there may be another factor involved as well. I have access to another computer running Arch Linux and the same versions of KDE (latest). A colleague and I tested similar configurations of user-places.xbel. The second computer does not show as steep a decline in Dolphin performance with increasing entries in user-places.xbel. The effect is still present, but much more moderate. I will continue experimenting to see if I can find this missing factor. I have some guesses: - Is it the icons used for folders? Maybe. - Is it the NFS mounts? No. We set up both computers with the same NFS mounts in /etc/fstab. - Is it the Dolphin startup directory? No. - Is it a hardware difference (such as CPU, RAM, SSD or GPU) between the test computers? No. - Is it recently-used.xbel? Apparently no (with simple testing). 3. There are no obvious differences in the straces or perf reports between the two computers. Obviously, we do not have debug symbols and we do not know how to do debugging or perf testing in great detail. One computer has much better Dolphin startup time and KDE file dialog performance than the other, but they do not have overall different performance on benchmarks or apps. (The slightly faster machine is the one with slower Dolphin & file dialogs). Questions: 1. We would like to copy versions of user-places.xbel from one computer to another for testing. How can we handle the unique ID in each entry in this file? What are those ID's matched against? Example: <ID>1580172204/1</ID> 2. Is there a way I can test Dolphin without any entries in user-places.xbel (without altering /etc/fstab)? 3. what other files, besides user-places.xbel and recently-used.xbel, could impact Dolphin startup times? Should we test with a specific dolphinrc configuration? Or should we delete that file for testing? Suggestions: Can you add an "ignore" feature (not simply "hide" the display of the entry, but more like "do not access") that would avoid placing unwanted system mount points in user-places.xbel? -- You are receiving this mail because: You are watching all bug changes.
