Niklas Hambüchen

Similar situation the other way around, where an alert did not get cleared for 20 days after it recovered: * https://github.com/AcalephStorage/consul-alerts/issues/275#issuecomment-3043174024

I can confirm this is still the case. My outputs also carry a timestamp, from which I can see that they are > 20 days old, `consul-alerts` has not fetched...

I found that restarting `consul-alerts` on one of the nodes clears all the old reminders, also those that were saved for completely different nodes. For example: `consul kv get -recurse...

Hm, even after the restart that sent the clearning slack notification, my log file at `/var/log/consul-alerts/consul-notifications.log` has the lines ``` [consul-notifier] 2025/07/06 23:32:09 Node=my-node-4 Service=myfs Check=Service 'myfs' check Status=warning [consul-notifier]...

I also hit this problem when building my app with the [nixpkgs](https://github.com/NixOS/nixpkgs/) package distribution. When upgrading from NixOS 24.05 to 25.05, which has the upgrade from 2.2.0 to 3.1.1 via...

Perhaps tangentially related: In nixpkgs we build both `.a` and `.so` using this hack: https://github.com/NixOS/nixpkgs/blob/55d1f923c480dadce40f5231feb472e81b0bab48/pkgs/by-name/li/libe57format/package.nix#L69-L84 ```nix # The build system by default builds ONLY static libraries, and with # `-DE57_BUILD_SHARED=ON`...

> 8 million files were not a target. Apparently to support them one has to make `getdents` syscall by hand which has not been done. Note that larger getdents buffers...

> Debugging shows me I'm always getting `d_type == DT_UNKNOWN` Use `strace` on `find` to determine the ultimate truth of what the syscall returns on your filesystem: ```sh mkdir -p...

Oh no. Good job finding that bug! A lot of code on the Internet seems to use `HAVE_DIRENT_D_TYPE`: https://grep.app/search?q=HAVE_DIRENT_D_TYPE That includes [cpython](https://github.com/python/cpython/blob/7afe1adb0089d0f2df2add79bbc2300c2f236d90/Modules/posixmodule.c#L15616) -- why does it work there? Looks like...

@TeofilC What exact error does it fix? Also, this does not include many of the changes of PR #157 -- would those changes also fix your problem?