Updated: December 1, 2021
Just recently, I tested MX-21 Wildflower KDE edition on my IdeaPad 3 machine. It was a blast. Among the best operating systems I've sampled in many an autumn. This great success buoyed my hope, and so I decided to expand my testing to the Xfce edition of this distro. To wit, the topic of today's experiment.
I will do the same here, again. Same hardware, same conditions. I believe it ought to be an interesting endeavor, because it lets us see how similar (or different) these two flavors are. If anything, my bittersweet Linux experience has taught me that variations among almost identical distro versions, e.g.: Kubuntu and Xubuntu or alike, can be vast. So I don't know what to expect here. Let's begin.
No luck this time!
Things started going badly right from the beginning. I booted into the live session. So far so good. I decided to make a bunch of visual changes, because things look small on the 1920x1080px 14-inch screen, a common problem with new hardware.
Immediately, I encountered a lot of problems. The process of making things pretty was much harder than what I accomplished in MX-19.3, and light years more difficult than doing it in KDE. But I will not go into these details just yet, because I don't have any screenshots. The reason being? My live session died!
How did this happen? Well, after about an hour worth of cosmetic work, I had to go for lunch. Working on battery power, I decided to suspend the laptop. When I came back and woke the machine up, the live session would not resume. I only got the wallpaper (with its purple flower motif), but no login screen. Nothing. No matter what I tried, the session would not resume. I was forced to reboot, and thus lost all my screenshots.
Un-very-fortunately, things didn't go well from here, either:
- Second boot, wallpaper only, no desktop - the boot sequence never progressed to a working session.
- Third boot, wallpaper only for 3 minutes, and then finally, the live session showed up.
I decided not to do anything but the installation itself for now. The process was similar to what I had in the KDE edition, and yet ever so slightly different. Consistency ftw! In the KDE one, the GRUB step offered me two ESP partitions, including the one on the live media, and it was the one (wrongly) selected by default. The Xfce one did not. There was only the one, correct on-hard-disk ESP marked. The whole thing took about four minutes, as before.
After reboot, the GRUB menu showed two identical MX-21 entries, and if you don't know which one is which, you cannot distinguish between the Xfce and KDE installations. The boot was not fast. In fact, it was very slow. Again, the system hanged in the post-boot-splash step, with the wallpaper shown for several minutes before I was actually able to do anything.
Plug in power, the screen dims. Predictable. P.S. You cannot up the brightness on the login screen, and when you combine that with the long login delay, it's really annoying.
The CPU was noisy. Constantly working. Not sure why. I then tried to run Firefox. It took several minutes to launch, eating processor cycles like mad. Eventually, it started, but it was odd.
The system wasn't very responsive. And there's a whole trace of crash errors in the system log, but at least you have none of that binary journalctl nonsense but plain, down-to-earth honest messages:
 Code: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 <00> 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 page:0000000044d2de81 refcount:1 mapcount:-1 mapping:0000000000000000 index:0x0 pfn:0x0
 flags: 0x7ffffc0001002(referenced|reserved|node=0|zone=0|lastcpupid=0x1fffff)
 raw: 0007ffffc0001002 ffffed4040000008 ffffed4040000008 0000000000000000
 raw: 0000000000000000 0000000000000000 00000001fffffffe 0000000000000000
 page dumped because: bad pte
[922.363846] CPU: 7 PID: 8964 Comm: IPC I/O Child Tainted: P OE 5.14.0-4mx-amd64 #1Debian 5.14.16-1~mx21+1
 Hardware name: LENOVO 81W0/LNVNB161216, BIOS E8CN23WW 07/28/2020
 Call Trace:
 ? __mod_lruvec_page_state+0x7f/0xa0
 ? do_epoll_wait+0xdb/0x760
 ? finish_wait+0x80/0x80
 RIP: 0033:0x7f830eb3a9b9
 Code: Unable to access opcode bytes at RIP 0x7f830eb3a98f.
 RSP: 002b:00007f8302bb2a88 EFLAGS: 00000246 ORIG_RAX: 00000000000000e8
 RAX: fffffffffffffffc RBX: 00000000ffffffff RCX: 00007f830eb3a9b9
 RDX: 0000000000000020 RSI: 00007f830e7c6080 RDI: 000000000000000c
 RBP: 00007f830e790880 R08: 0000000000000001 R09: 000000007fffffff
 R10: 00000000ffffffff R11: 0000000000000246 R12: 0000000000000000
 R13: 00007f8302bb2cd0 R14: 00007f830e71bb00 R15: 00007f830e720c00
This looks like a problem that the AHS kernel should address. My wild guess, this edition wasn't baked correctly, hence a whole slew of hardware-related process, including all of what we've seen so far.
Now, if you read my Linux kernel crash analysis book, you will find the analysis quite interesting. With the CS register dump pointing to 0033 (odd last digit, details in the book), this means kernel oops triggered by a userland error. That is normally not possible, unless you have a hardware error or a bad system call. Since the laptop is stable and works well with the other two operating systems, including MX itself (and Windows 11), my guess is that this version has a wonky platform-specific module that does not cooperate well with IdeaPad's firmware.
Furthermore, during the boot, you can see an AMDGPU error show on the screen, which is NOT there when you boot the KDE one. Hint, this would also most likely explain why Firefox isn't launching, especially if it expects to use hardware acceleration (enabled by default), and there ain't no proper firmware support, so the system can't use the GPU correctly, hence the CPU revving. So this looks like a really bad issue, and it kills my entire review.
Other than that ...
Well, the visual styling was bad. Worse than in MX-19.3. So, increasing the scaling with xrandr just makes everything blurry, like blowing up a picture above its native resolution. Conky also disappears from the screen when you do this. A pretty bad design.
Font scaling and font size increase, at least, work fine. But making the bottom panel taller led to tons of problems. The PulseAudio icon and the battery indicator icon behave separately from the other indicators in the system area. While you can resize the rest (automatically or manually), these behave on their own, so this looks ugly. The Bluetooth icon also loses its nice style and becomes the ancient Gnome-2-like 1999 icon.
I also had to make the digital clock bigger. There is a dock-like applet for the panel, but you cannot manually rearrange icons there, so you need to unpin and then pin icons to get them sorted as you like. When you combine all these unnecessary woes with the hardware issues, we have a dud at our hands.
MX Linux MX-21 Xfce is the complete opposite of my MX-21 KDE review - that one was delightful. The Xfce one is the worst experience I had with this distro, probably ever. I didn't really get to properly test anything due to the general sluggishness, the login freeze, the suspend & wake problems, the Firefox slowness, the kernel oops, and all the rest of it. But the visual customization did show me one important aspect - how much more advanced KDE is, and how fragile scaling is in Xfce.
I really am not in the mood to manually tweak 20-30 separate Xfce elements just to have a nice, presentable desktop. That's 2005, and it needs to stop. The Xfce version of MX-21 ain't bad, but it's fragile. Worse yet, the distro behaved far better in the past, so we also have a regression on our hands. All I can say, go for the KDE version, it's amazing (among the best systems I ever tried). Whereas the Xfce one needs to go back to the workshop and get some serious rework. Alas, on that note, and with some mild paranoia swirling in my brain, we end this sad review.