https://bugs.kde.org/show_bug.cgi?id=398346
k...@happyjack.org changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |k...@happyjack.org
--- Comment #3 from k...@happyjack.org ---
I had the same problem, and have found the source.
In ksysguard, in ksysguardd/Linux/cpuinfo.c, there is a buffer meant to hold
/proc/cpuinfo. It has a fixed size of 32 * 1024.
This size is too small for some modern, multi-core systems. I have a 24-core
system, and each entry in /proc/cpuinfo takes up around 1500 bytes. My entire
/proc/cpuinfo is 35982 bytes, meaning it won't fit into the buffer, and as a
result, ksysguardd doesn't register monitors like system/cores, which causes
the system load viewer to fail.
I don't know the right size for the buffer, but I'd say "significantly larger"
is proper. I can't imagine (for example) 128K being onerous at all, and it'd
allow for significantly larger /proc/cpuinfo files. But that might not even
work, given that apparently AMD is coming out with 64-core/128-thread
systems... It should probably just be a megabyte or something, so you never
have to worry agin. Or just allocate dynamically, with a gigantic ceiling.
Anyway, the fix is to increase the size of the buffer one way or another.
--
You are receiving this mail because:
You are watching all bug changes.
