>> Teal deer: don't skip the warning. (But maybe don't print it unless
>> nroff is taking longer than 5 seconds?)
> Yeah, don't try nroff, but the mandoc we are actually using.
As far as I know I don't have access to a slow machine running anything
recent enough to have mandoc, so I can't give actual times.
But I have access to a 9.1 amd64 machine. Using mandoc there, I have
trouble timing because that hardware is so fast. But complexity still
makes a significant difference. Using time(1), with the same seven
files copied from 1.4T (so as to change less between tests), and
running each one 100 times to get a couple more significant digits:
for i in *;
do
ls -l $i
time sh -c "for a in 0 1 2 3 4 5 6 7 8 9; do
for b in 0 1 2 3 4 5 6 7 8; do
mandoc < $i > /dev/null
done
done"
done
=>
-r--r--r-- 1 mouse wheel 109360 Jun 27 2024 gcc.1
2.68 real 2.22 user 0.46 sys
-r--r--r-- 1 mouse wheel 140179 Jun 27 2024 ksh.1
2.26 real 1.96 user 0.28 sys
-r--r--r-- 1 mouse wheel 104862 Jun 27 2024 lex.1
2.19 real 1.91 user 0.28 sys
-r--r--r-- 1 mouse wheel 69334 Jun 27 2024 mdoc.samples.7
2.75 real 2.26 user 0.48 sys
-rw-r--r-- 1 mouse wheel 134421 Dec 31 2024 moussh.1
3.74 real 3.33 user 0.41 sys
So moussh.1, a little smaller than ksh.1, takes over half again as much
time; mdoc.samples.7, less than half the size of ksh.1, takes over 20%
longer. So, using the largest manpage is likely to give you a
misleading result.
A rough comparison against nroff on the 9.1 amd64 machine indicates
nroff is slower by a factor of about 13, which converts the 26 seconds
from the SS20 into 2 seconds. This is close enough to your 5 seconds
that I find it plausible that a slow machine could take that long. I'm
fairly sure a low-end hp300, mac68k, or vax would be at least a factor
of 2.5 slower than the SS20. (Likely others, too; those are just the
ones I am confident of.)
/~\ The ASCII Mouse
\ / Ribbon Campaign
X Against HTML [email protected]
/ \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B
