[Jackit-devel] RE: [PlanetCCRMA] jack memory leak?

[Paul Davis]

>   I left the machine with jack running overnight. The memory usage (as
>displayed by top) this morning (7AM) is now 260556K, up from about 85000K
>yesterday when I left the office at 4PM. No other 'applications' are
>running. Obviously lots of other 'stuff' is running.

using top to determine memory usage is an apparently sensible but
actually incorrect thing to do. free(1) is a better program.

here is the state of my system right now:

paul[2092]>cat /proc/meminfo 
        total:    used:    free:  shared: buffers:  cached:
Mem:  923262976 781770752 141492224        0 40984576 568291328
Swap: 139821056  6516736 133304320
MemTotal:       901624 kB
MemFree:        138176 kB
MemShared:           0 kB
Buffers:         40024 kB
Cached:         553456 kB
SwapCached:       1516 kB
Active:         475764 kB
Inactive:       161272 kB
HighTotal:           0 kB
HighFree:            0 kB
LowTotal:       901624 kB
LowFree:        138176 kB
SwapTotal:      136544 kB
SwapFree:       130180 kB

paul[2093]>free -b           
             total       used       free     shared    buffers     cached
Mem:     923262976  781778944  141484032          0   40984576  566738944
-/+ buffers/cache:  174055424  749207552
Swap:    139821056    6516736  133304320

so, if you look at the line from free(1) that starts "-/+", what i
think this means is that even though we are apparently using 781778944
bytes, only 174055424 is actually in use by programs and the core part
of the kernel. the actual "free" memory is 749207552, most of which is
being used by the kernel's disk buffers and cache (actual total of
that usage is 40984576+566738944 bytes.

does that help?

--p

ps. i've had a jackd running for four days now, through a whole series
of sometimes segfaulting sessions with ardour. it feels much more
stable after the last round of changes that it used to be.