Duje Mihanović's Site

Fixing a trash-picked Acer Aspire E1-731

2025-06-11T17:23:47Z

Recently I found an Acer E1-731 on a random junk pile. At a first glance it seemed relatively modern (which I concluded from the presence of USB 3.0), but also heavily water damaged. While most would conclude that it cannot be saved at that point, I decided to take it with me anyway to see if there’s anything worth salvaging and try to repair the laptop itself.

First impressions

The first thing I did upon taking the laptop was to remove the main battery, as any current would massively speed up corrosion of the board. Once the laptop was at my workstation, I took off the bottom cover and found that what had been exposed of the motherboard was unsurprisingly rather dirty, but there was no corrosion to be found.

With the back cover gone I was able to remove the 4GB DDR3 SODIMM, 500GB WD Blue hard drive and an Atheros WiFi card, which were then dried and the SODIMM and WiFi further cleaned up with IPA and again dried. With these aside, I continued disassembling the laptop to dry everything as quickly as possible and also to remove the RTC battery (for the same reason as the main battery). Not coincidentally, there were a few small corroded areas around the RTC battery socket, but besides that the board was in pretty good shape.

With all of the parts dried and therefore safe from any further corrosion, I decided to test the salvaged RAM and HDD. Both ended up working and the HDD is also perfectly healthy (judging by the reallocated sector count, which is 0), which alone made picking up this laptop worthwhile. The main objective however was to repair the laptop itself, so then I proceeded to have a look at the motherboard itself.

Diagnosing the board

After having made sure that the board was dry and clean of corrosion, I detached the DC in jack and power button from the bottom case and hooked them up to the motherboard to give it a test boot. However, it gave no signs of life: no LED lit up at all and pressing the power button also did nothing.

My first action was to measure the voltage at the power button header and I found that it was abnormally low (~300-400mV for a signal meant to be 3.3V). After some more probing around I found out that the main rail was also gone, which I traced down to the DC in circuit. To be more specific, it seemed that something was wrong with the DC in MOSFET as its drain was at the expected ~20V, but the source at just ~900mV.

At this point, I needed schematics and a boardview, as the board itself is devoid of any component markings with the exception of screw holes and connectors. Fortunately, they were rather easy to find (“VA70/VG70 rev 2.1 schematic boardview”) and so I proceeded to ask and search around to see what could be going wrong. I first checked the downstream VRMs for any shorts or overloads which were said to potentially cause this, but I found nothing. Next, I went through the charge controller circuit as suggested by a certain forum post, but also found nothing wrong there.

Eventually, I had a look at the 2nd DC in MOSFET (Q8802), right between the 1st one (Q8800) and a battery charging MOSFET. I have found that the R_DS(off) of this MOSFET was just about ~500Ω, which I suspected to be extremely low considering a enhancement-mode MOSFET is not meant to conduct at all while off. I desoldered this MOSFET and when powering the board again, there was ~20V on both the drain and source of the 1st DC in MOSFET.

From that and the apparent fact that nothing was wrong downstream, I concluded that the 2nd DC in MOSFET was faulty and needed a replacement.

Sourcing a replacement MOSFET

I decided to buy the replacements from AliExpress as usual. There is, of course, a risk of getting fakes, but in this case the MOSFETs I bought are genuine. Specifically, the MOSFET in question is an IRFHS8342, and I bought the replacements here.

Replacing the MOSFET

After the MOSFETs arrived, I soldered one onto the board. The specific procedure I used was to tin the pads with my iron in advance and apply flux liberally before applying hot air.

When I plugged in the board (after letting it cool down, of course), it would click at a regular interval and otherwise show no signs of life. This was still progress, so while remaining cautiously optimistic I made a few more measurements around the board. The main rail voltage had now gone up to oscillating at around 4V, which could suggest that there’s a VRM being overloaded, but just like the previous time I found no shorts anywhere.

While desoldering the dead MOSFET, I had accidentally knocked two capacitors between it and the other MOSFET (C8802 and C8803) off the board. Just to rule out these two causing issues, I went ahead with soldering these two back on. C8802 was soldered on easily, but on the other hand I dropped and lost C8803 while trying to solder it back on. Nevertheless, with C8802 again present the board finally powered up!

For reference, here’s these 2 capacitors in the schematic:

With the board finally showing signs of life, the only thing remaining to do was to check whether it outputs an image. I initially tried HDMI, but there was no output at all. On the other hand, with the internal display connected it POSTed straight away. From here it was simply a matter of installing a new RTC battery and reassembling the laptop.

Final words

The laptop now runs well, but not without flaws. The water had sadly not been kind to the display and as a result the light guide is visibly distorted, which doesn’t really have an impact on usage but isn’t nice to see either.

The main battery was found to be dangerously discharged (~1.4V per cell), however the BMS successfully revived it and reports a full charge capacity of ~30Wh with a design capacity of 48.4Wh, leading to a wear level of around 39%. A battery calibration together with time will reveal whether this figure is true, but I am nevertheless exceedingly happy to have a working battery.

Finally, I believe that the clicking thing serves to confirm the importance of replacing dead filter capacitors. Even if the board booted without C8802, it’s not impossible that there would have been stability issues later on. I have since found C8803 again, however I will probably not disassemble the whole laptop again just to solder it back on. If I do end up tearing it down again for another reason or the laptop shows any instability, I will take the time to solder it back on.

Configuring Git over HTTP(S) with Caddy

2025-05-17T17:16:22Z

Since I switched to Caddy from Nginx, the only thing that worked on Nginx but I just couldn’t get working on Caddy was cloning Git repositories over HTTP.

I had initially tried sort of backporting the Nginx configuration, which didn’t work. After some time I ran into James Atkins' post in which he detailed how he tackled this issue. Again, this didn’t work for me (possibly because I use GitWeb rather than cgit), so I ended up just configuring a Git daemon as a workaround until I eventually figure it out.

That has finally happened, so here I’ll share my working configuration and try to explain it.

The configuration

First, let’s see the configuration itself:

@git_cgi path_regexp "^.*/(HEAD|info/refs|objects/info/[^/]+|git-upload-pack|git-receive-pack)$"
@git_static path_regexp "^.*/objects/([0-9a-f]{2}/[0-9a-f]{38}|pack/pack-[0-9a-f]{40}\.(pack|idx))$"
@push-info query service=git-receive-pack
@push path_regexp "^.*/git-receive-pack$"
root / /usr/share/gitweb
try_files {uri} gitweb.cgi
handle @git_cgi {
reverse_proxy unix//run/fcgiwrap.socket {
transport fastcgi {
env SCRIPT_FILENAME /usr/lib/git-core/git-http-backend
env GIT_PROJECT_ROOT /var/lib/git
env REQUEST_METHOD {method}
env QUERY_STRING {query}
env PATH_INFO {path}
}
}
}
handle @git_static {
file_server {
root /var/lib/git
}
}
basicauth @push bcrypt restricted {
}
basicauth @push-info bcrypt restricted {
}
reverse_proxy unix//run/fcgiwrap.socket {
transport fastcgi {
env GITWEB_CONFIG /etc/gitweb.conf
split .cgi
}
}
handle /static/* {
file_server {
root /usr/share/gitweb
}
}

Explanation

Matchers

The first two matchers (@git_{cgi,static}) are for repository data, of which the first matcher is handled by git-http-backend while the other is just plain old files which are served by Caddy for performance reasons.

The other two are for pushing over HTTP, which will be covered later.

`root` and `try_files`

My root directive may seem unusual at first since it matches / instead of *, however this was necessary because otherwise all requests, including the ones covered by the @git_* matchers ended up being handled by GitWeb with predictable results.

The try_files directive simply uses the GitWeb CGI script as the index document.

`reverse_proxy`

The “root” reverse_proxy is almost not worth explaining; it simply makes Caddy run GitWeb through FastCGI.

The various `handle`s

This is the interesting part.

The /static/* handle takes care of GitWeb’s static files, which include a JavaScript file, a stylesheet and some (fav)icons.

The @git_cgi handle is probably the most important one and the next troublemaker after root: after resolving root any git clone would result in a HTTP 500 (actually, not cloning, but curling <repo>/info/refs). The gist of it is that git-http-backend relies on a number of environment values which do not seem to be filled in by Caddy, so it had to be done manually (for some reason, apparently only I seem to have to do this).

Figuring out what had to be added was puzzling, as FastCGI did not seem to have been logging in either the journal or somewhere in /var/log, so I ended up manually running git-http-backend with the aforementioned environment variables and that way I could know why the backend was failing. From there it was just a matter of adding the required environment variables to make the backend happy.

Particularly worth elaborating are {path} and {query}. git clone will issue a GET request on /<repo>.git/info/refs?service=git-upload-pack, and the backend does not like having the service query in its PATH_INFO, which would happen if {uri} was used instead. The query is still needed for detecting v2 protocol support, so it is passed to the backend separately in QUERY_STRING. git-http-backend(1) has more information on all of this.

Besides that, I also had to juggle around with permissions on the Git root, first to make the backend work at all and then to make pushing work; I eventually settled on git:www-data, 775 (on directories) and sudo -u www-data git config --global safe.directory '*' (user names might differ between distros, mine is Debian 12).

`basicauth`

They add a layer of authentication on authenticated pushes. As git-http-backend(1) suggests, I have added authentication to both ref advertisement and the call to git-receive-pack.

If you don’t need/want/whatever HTTP pushing, you could drop these directives and the two push matchers. I’d also recommend dropping |git-receive-pack from the @git_cgi regexp to make sure it may not be called.

Note: the directive is basic_auth on newer Caddy versions (2.6.2 is currently packaged in Debian 12). Keep this in mind if Caddy errors out here for you.

Final words

While this configuration works fine, I suspect it may be optimized (read: shortened) further. In particular, I suspect the push matchers and their respective basicauth directives could be squashed together, but I can’t really think of a way to do that (or any other potential optimization).

If you have any ideas on this or questions in general, feel free to use the reply button below or contact me via another medium.

Fixing the scroll wheel on a SteelSeries Rival 310

2024-09-14T19:47:16Z

If you’ve been using a computer mouse for a while, you might get an issue where the scroll wheel moves randomly, jumps around while scrolling or at times doesn’t respond at all. My mouse also developed this issue recently, but luckily it’s very easy to fix. The only thing that happens is various gunk builds up on the wheel sensor and this causes misreadings, so obviously cleaning up that gunk makes the wheel work like new again. Here I’ll show how I did that on my mouse, but the general idea applies to any mouse.

What you need

No matter what, you’ll need the following:

A mouse
Corresponding screwdriver(s) - In 99% (if not 100%) of cases a Phillips #1 will be enough
Some way to remove the gunk: toothbrush and/or blowing
A working brain cell

If the screws are hidden under the mouse feet (like in my case), you might also want to have a hairdryer or hot air station together with a sharp object to remove the feet without damaging them. To do this, heat up the feet (on low heat, obviously; I used 100C as that’s the least my hot air station can do) until you can no longer touch them, then pry them away with the sharp object. I don’t think I did the best job at this, but I was able to reapply them acceptably well after the repair.

Fixing the mouse

First, for reference, I’ll throw in a picture of the mouse:

After removing the feet, the screws become visible:

Here you can see why I said I didn’t do the best job of removing the feet: some glue remains on the mouse. I’m quite sure that after a perfect removal it would all be on the feet and this would allow them to be reapplied perfectly, but as I said I was able to do so well enough IMO.

After unscrewing the 4 screws revealed under the feet, the bottom and top halves of the mouse come apart easily. From there I was able to take a look at the scroll wheel assembly, and there sure was something to see:

To get to the scroll wheel in this case, I had to unscrew that LED BOARD, take it out of the wheel (which took a few seconds) and then unscrew the entire mainboard from the chassis. Once that is done, I attacked both the wheel and the sensor with a toothbrush and blown air and then gave the rest of the mouse a much needed clean while at it. This picture of the area under the board gives a good idea of how dirty the mouse was:

Once I cleaned that up somewhat, putting the 2 boards and then 2 halves of the chassis back in place was pretty easy. To test the mouse, I screwed in 2 of the chassis screws diagonally, plugged the mouse into my PC and hoped everything would work, which it did. After the other 2 screws came reapplying the feet; because I had not removed them perfectly, they didn’t seem to pop back in well. I worked around this by pressing them against with the chassis with the sharp object, and as I have already said this gave an acceptable result.

Conclusion

With that, the mouse was back in full order. One issue I didn’t fix yet is the side panels ungluing after a liquid accident long ago, but I doubt that’s worth describing.

I hope that in the future I will be able to write about repairs of more complex devices.

Going offline

2024-08-27T13:41:18Z

dujemihanovic.xyz will be going offline today, August 27th, 2024, sometime around noon or afternoon according to UTC and will remain offline for a bit more than a week.

An explanation will be posted after the site goes back online.

Update: The shutdown has been called off.

Now Available: Skeleton Theme

2024-08-25T12:30:02Z

The barebones theme used by my website has been split up into its own repository. If you’re interested, check it out here and an example commit to actually use it here.

Note that I haven’t thoroughly split them up quite yet. This means there may be remaining mentions of my name in the theme. I expect to correct that sometime.

Software Review: FreshRSS

2024-08-21T12:27:03Z

Most RSS users start off with desktop readers such as Newsboat or Akregator in order to keep things simple. However, due to certain drawbacks of this approach (which I will talk about below) I have looked for an aggregator I can install on my VPS and use as such. After a bit of research, I have decided to go with FreshRSS and here I’ll explain why I decided to ditch Akregator in the first place and why I chose FreshRSS.

What even is RSS?

If you’re already initiated on this subject, feel free to skip to the next section.

A practically universal use case for the Internet is reading various blogs and news. However, as your number of followed blog and news sites builds up, going through all of them looking for new content becomes a serious chore. To make things worse, many (generally news) sites are becoming increasingly bloated. (Not just technologically, but also with garbage content!)

RSS solves these issues beautifully. The idea is that each site has an XML document (called an RSS feed, or feed for short) which contains a list of the N latest posts on that website. You can add an arbitrary number of these feeds into your reader and it will automatically sync these at a specified interval, optionally notifying you of new content. This removes the chore of having to go through your followed sites manually, but it also works around the technological bloat issue as the feeds can only contain plain text or (more common and still reasonable) plain HTML. Finally, low-quality content can be much more easily ignored or straight up filtered from your reader.

Unfortunately, RSS has declined since its golden age in the 2000’s with Firefox dropping its own reader and many people forgetting about RSS, however it persists on many sites (even big ones) and other readers continue to be actively developed.

What’s the problem with traditional readers?

If you consume your feeds on multiple devices (as most people, me included, probably do) with a traditional reader, your feeds and other info on them (such as read and important markers) will not sync. Besides that, you might lose articles on high-volume feeds if you do not open your feed reader for a while (which might be just a few days) and the feed has a sufficiently low N.

The solution to these problems is to run your aggregator on a server. There exist public ones, but a far better solution would be to host your own. This is where FreshRSS comes in.

How to install it?

If you have Docker Compose, put something like this into a docker-compose.yml:

version: "3.8"

volumes:
 data:
 extensions:

services:
 freshrss:
 image: freshrss/freshrss:latest
 container_name: freshrss
 hostname: freshrss
 restart: unless-stopped
 ports:
 - "8181:80"
 logging:
 options:
 max-size: 10m
 volumes:
 - data:/var/www/FreshRSS/data
 - extensions:/var/www/FreshRSS/extensions
 environment:
 TZ: Europe/Zagreb
 CRON_MIN: '3,33'

Then, configure your web server to point a subdomain or path to the respective port. Open the subdomain or path in your browser, do the initial setup and you’re done. Add a few categories and feeds, then go through the settings and configure everything to your wishes.

Since FreshRSS is a PWA, you may “install” it using a Chromium-based browser (which I do on Android) or through the Web Apps program on a Linux Mint system.

How does it feel to use? Should I use it?

There’s not that much to say. Adding feeds is as simple as on a traditional reader, all my feeds and starred articles sync between devices (strictly speaking, they do not, but are fetched from the server) and it overall behaves pretty much just like Akregator did excluding the issues.

If you only read feeds on one device and don’t care about losing articles from big news sites here and there, you should probably stick with a traditional reader as that is obviously less complicated, probably a bit more secure, allows offline reading and so on. However, if your RSS experience is affected by one of the aforementioned issues, I can highly recommend FreshRSS as it’s relatively simple to set up and does not otherwise try to be radically different just for the sake of it.

Matrix delegation and how it may bite you

2024-08-21T12:27:03Z

For those who don’t know, delegation in Matrix is used in server-to-server communication to figure out which server serves a given domain. As an example, if my own Matrix homeserver was running on matrix.dujemihanovic.xyz instead of dujemihanovic.xyz, I could delegate the latter to the former to save anyone wanting to contact me from having to type out the matrix..

Besides domain name, delegation can also be used to specify which port to use for server-to-server communication. The default is 8448, and if it’s blocked you can use delegation to use 443 for server-to-server as client-to-server does by default. However, if you can, I’d strongly suggest using 8448! I had been delegating S2S to 443 almost the whole time I have had this server for no reason and it seems that this caused an extremely weird issue with a certain room:

What happened?

Message fetching kept breaking constantly. What I mean by that is that when I joined the room everything would work fine the first few messages, but at some point I would start getting notifications without any new message being present in that room. I have noticed that logging out and back in would get the missing messages in my client, but then the forementioned cycle would repeat again no matter how many times I logged out and back in (this also happened on other clients besides Element desktop). To confirm my homeserver was the issue, I joined the room with my old matrix.org account and sure enough that worked just fine.

I tried the usual things such as restarting Dendrite and the whole VPS, but to no avail. I was pretty insistent that the issue was not with my homeserver but the main server hosting the room (which, unsurprisingly, turned out to be false) and so I gave up on that. The eyeopening moment was me reading the conduit documentation (I had considered migrating to it), specifically this:

If Conduit runs behind Cloudflare reverse proxy, which doesn’t support port 8448 on free plans,

This implies that routing server-to-server traffic to 443 should only be done if it’s absolutely impossible to use 8448 for this, and the Synapse documentation said something similar:

However, if your homeserver’s APIs aren’t accessible on port 8448 and on the domain server_name points to, you will need to let other servers know how to find it using delegation.

Fixing the issue

Encouraged by this, I fixed up my server:

allow port 8448 in ufw
add something like this to Caddyfile:

dujemihanovic.xyz:8448 {
reverse_proxy /_matrix/* localhost:8008
}

change /.well-known/matrix/server to point to dujemihanovic.xyz:8448 (in theory, I could have gotten rid of that return directive altogether as 8448 is default anyway, but I still chose to specify it just to be safe)
reload caddy and restart dendrite (the latter is, again, just to be safe)

Once all this was done, the room finally started acting normally.

Small sidenote

I must note that delegating federation to 443 should not cause breakage like this. Despite this, it still did so in my case and for that reason I wrote about it anyway. It’s very unlikely that you will be affected by this issue, but I still believe it should be pointed out in the event that it does.

My projects are now listed on the website

2024-08-21T12:27:03Z

I have created a new section of the site dedicated to my projects. Currently the section numbers one whole project, that being my effort to get Marvell PXA1908 booting mainline Linux. Check it out, let me know if you have any suggestions for the new page and, if you can, contribute to the effort!

RSS updates

2024-08-21T12:27:03Z

I have made some small, but important changes to the way the RSS feed of the site is generated.

The biggest change is the relocation of the feed from dujemihanovic.xyz/posts/index.xml to dujemihanovic.xyz/index.xml. To make sure subscribers (assuming any exist) don’t get cut off by this change, I have set up a 301 redirect from the old location to the new one and I don’t plan to remove this redirect, although it probably won’t hurt to change the feed URL anyway (Akregator and Newsboat, the readers I tested, don’t seem to care much about the 301 and keep using the old URL).

I have also done some small cleanup of the code, the commit range in the repository is 5a43568..47271c2 for those interested. I have also looked into adding an Atom feed at some point.

Updating/installing Dendrite

2024-08-21T12:27:03Z

I decided to share the method I just devised for updating (and perhaps installing) Dendrite on my VPS.

Motivation

I discovered that yesterday (as of writing), Dendrite was updated to 0.13.2, and I was running 0.13.1. Because old thing bad and new thing good, I decided to update right away!

The method I tried using for this was essentially the same as I used for the initial install:

Stop dendrite service
Download new tarball
Unpack new tarball into existing dendrite directory
Run go build -o ./bin ./cmd/...

This is where the issues start: it errored out on some nats something and I assumed it just didn’t like being unpacked into an existing dendrite directory, so I made a new one. It failed again and then I read the actual error that occurred, and as it turned out the version of Go packaged in Debian is too old to compile the nats thing. Isn’t that a classic?

I looked around to see if maybe Go has its own Debian repository, but I was out of luck. It seemed that the 2 options were either install Go from source or just stick with Dendrite 0.13.1, but in the end I decided on a 3rd solution:

Compiling it locally

My personal machine runs Artix, so there should be no problem with outdated Go (or anything really) whatsoever. I figured that I could compile Dendrite on my personal machine and then rsync the binaries onto my VPS, just like I did with my Piped-Material frontend. Here are the steps:

Clone Dendrite git repository
Checkout v0.13.2 (replace with latest version)
Run go build -o ./bin ./cmd/... and wait for it to compile
On VPS, move /opt/dendrite to /opt/dendrite-old and create new /opt/dendrite
(optional) Strip the binaries so they take up less space: strip --strip-unneeded bin/*
rsync the bin directory: rsync -aAXvhzP bin root@vps:/opt/dendrite/
Most likely all these options aren’t needed, it is what I use for VPS backups so I decided to not touch it just to be safe.
Back on the VPS, copy over matrix_key.pem, media_store, (if it exists) searchindex and dendrite.yaml from dendrite-old into dendrite.
I have noticed that Dendrite also creates a jetstream directory, I’m not sure what it is for but my homeserver seems to work fine without copying it over.
Update configuration file. To do this, I went back to the Git repository on my local machine and ran git log -p dendrite-sample.yaml. This time the only change was that MSC2946 (Space Summaries) was de-MSCified, so I removed that from the config file.
Make sure the user and group dendrite own every file in /opt/dendrite. Run: chown -R dendrite:dendrite /opt/dendrite

After doing this, I started Dendrite again using systemctl and to my surprise the homeserver was up and running first try. I also took this time to enable server-side search and DNS caching, which is something I overlooked during the initial setup of my homeserver. I have also verified that space previews still work despite MSC2946 being removed from the configuration file.

This way I was also able to save a bit of disk space, which is important since space is often rather limited on VPSes.

Dropping Gitea

2024-08-21T12:27:03Z

I have decided to drop Gitea as it’s only really useful when you have multiple people collaborating on a project and is just bloatware otherwise. To demonstrate, here’s how much bandwidth it takes to load in the old Gitea site’s main page:

Here’s how much it takes up with the amazing gitweb:

If you do the math, you will find that the new gitweb site is roughly 22.18 times smaller than the Gitea site! This will certainly help with browsing on older computers and weaker internet connections.

(Also notice that gitweb does not make that pesky 3rd-party request!)

I also expect to move some of my previous projects from Git{Hub,Lab} here relatively soon.

What about the old site?

I will keep it as an option in case I by some miracle end up creating a project with lots of outside contribution as that’s when it would become truly useful. Until then, I always accept patches sent with the venerable git send-email.

If the new site breaks down in any way, please let me know.

Windows Backup fails with 0x81000037? Check your filesystem

2024-08-21T12:27:03Z

The Problem

Yesterday, in preparation for a “downgrade” to Windows 10 from 11, I was trying to make a backup of my Windows files (using the Windows 7 Backup and Restore control panel). Not long after it actually started copying the files, it spat out this error:

Windows Backup failed while trying to read from the shadow copy on one of the volumes being backed up. Please check in the event logs for any relevant errors.

The error code listed was 0x81000037, and the first thing I did out of instinct was look this code up. The official Microsoft support topic stated that this error is caused if a reparse point (Windows slang for symlink) points to a FAT volume or is located on a volume containing compressed files. The listed workaround for this error is to delete the offending reparse point, and to locate it the topic suggests using the command dir /al /s, which is roughly equivalent to find -type l in Unix-like systems.

The first roadblock occurred when I found out that there is no offending reparse point which could cause this error, so I consulted eventvwr and found out that Windows Backup leaves some logs at %WINDIR%\Logs\WindowsBackup. It seemed to me that there are certain cache files that caused it to fail, so I deleted them and tried again. The backup again failed and I tried doing this a few more times, but as nothing helped I eventually gave up.

The Solution

Having been unable to fix Windows’ backup utility, I decided to just copy the files to the backup drive and copy them over back after the “downgrade”. This is when the actual solution to my problem becomes visible: Windows Explorer errored, I forgot the exact message by now but it went something like “The file XYZ is not available at the location” or something, searching for the exact message is as pointless as always.

This usually indicates filesystem errors, so I ran chkdsk C: and, unsurprisingly, it found errors on the filesystem. Besides others, it mentions the files XYZ that couldn’t be copied to the backup drive. I ran chkdsk /scan C:, rebooted through the filesystem error notification that popped up right after (chkdsk /f C: and reboot will work as well) and let it fix the errors. Once it rebooted, the backup completed without errors. After installing Windows 10, restoring the backup was a breeze (I only had AppData to back up, everything else was already on other drives) and so I was quickly up and running, then amazed by how fast and snappy Windows 10 is compared to its “successor”, but that’s worthy of a separate post.

(I don’t expect to get around to writing such post anytime soon though.)

“Why are you writing about this?”

As you may know, Microsoft stuff is notorious for being not very well documented and it’s entirely possible (even if very unlikely) that someone else encounters the same issue I have and is left helpless thanks to Windows’ lack of a proper, easy-to-use, centralized and well-written technical documentation system (think of man’s sections 4, 5, 7 and 8 for an idea of what I mean), not to mention countless YouTubers that parrot the same solutions as mentioned on the web many times before. I hope that this article will be useful to the one or so person with the same issue that may stumble upon it.

TLDR: Nobody seems to have mentioned this before (please correct me if I’m wrong), so might as well help the potential other person stuck in the same boat as me.

Site is now on Gitea

2024-08-21T12:27:03Z

Even though the site is still nothing special, I have published its source code on my new gitweb instance. To save you a click or two, the repository itself can be found here. If you have any suggestions as to how to improve the site, feel free to contact me. If you are willing to directly contribute, I also accept patches sent with git send-email, a great tutorial on how to use it can be found here.

For convenience and to emphasize this, I have also added a “This site is free software” link to the footer, right under the RSS button.

Edits

2023-08-19: Change links to point to gitweb instance

First Post

2024-08-21T12:27:03Z

This is the first post on my new blog. I don’t expect to write much at all, but if I do it will most likely be about my experiences with Linux on smartphones and perhaps reviews/recommendations of products I come across.

I’m still experimenting with Hugo, so don’t be surprised if the site layout changes dramatically in a short period of time. I put the source code for this site under a (as of writing, local) Git repository which I do intend to make public if and when the blog becomes something relatively serious.

EDIT: The source code is now public: Site is now on Gitea

Duje Mihanović's Site

Fixing a trash-picked Acer Aspire E1-731

First impressions

Diagnosing the board

Sourcing a replacement MOSFET

Replacing the MOSFET

Final words

Configuring Git over HTTP(S) with Caddy

The configuration

Explanation

Matchers

root and try_files

reverse_proxy

The various handles

basicauth

Final words

Fixing the scroll wheel on a SteelSeries Rival 310

What you need

Fixing the mouse

Conclusion

Going offline

Now Available: Skeleton Theme

Software Review: FreshRSS

What even is RSS?

What’s the problem with traditional readers?

How to install it?

How does it feel to use? Should I use it?

Matrix delegation and how it may bite you

What happened?

Fixing the issue

Small sidenote

My projects are now listed on the website

RSS updates

Updating/installing Dendrite

Motivation

Compiling it locally

Dropping Gitea

What about the old site?

Windows Backup fails with 0x81000037? Check your filesystem

The Problem

The Solution

“Why are you writing about this?”

Site is now on Gitea

Edits

First Post

`root` and `try_files`

`reverse_proxy`

The various `handle`s

`basicauth`