Re: [PATCH 3/3] readutmp: load libsystemd with dlopen instead of linking

On 2026-07-19 10:50, Luca Boccassi wrote:
> the system just breaks
> otherwise (as in, the installation cannot proceed, full stop).
Yes, that's the static check I was referring to. And for coreutils this 
particular static check results in a false positive for libsystemd, as 
coreutils works well enough even if the static check is omitted.

As I understand it, the current effort in Gnulib is how to pacify the 
false positive in a way that doesn't require intrusive changes to 
working source code by adding dlsym calls all over the place.


> this is not something that can or should be changed in any way.

I don't see why not. GCC developers over the years have developed many 
ways for applications to suppress or bypass GCC's false positives. (I 
added some of those ways myself, decades ago.) This effort has 
encouraged developers to put GCC's sometimes-fallible static checking to 
good use. I imagine that Debian's packaging system could do likewise 
without too much trouble, as an extension to how it currently behaves. 
This would be more productive overall.