david@elroy.jpl.nasa.gov (David Robinson) (07/10/89)
Henry pointed out that there is not much you can do to newsrun to
speed it up without breaking it's robustness. I agree, but you
can tweek it a bit for added performence. The first and biggest is
to comment out the c7decode lines if you never expect to get encoded
news. The other thing I have done which helps if you get many small
batches (ie from a once a minute NNTP sender) is to rewrite 'spacefor'
and 'sizeof' into C if you can. Both are trivial to write but you
may need a statfs() system call for the former. This reduces the
numerous ls, df, and awk calls.
I do agree that C news is considerably faster, our Sun-2 had just stepped
over the brink and was not able to keep up with the volume, with C news
it has been given a reprise.
-David
--
David Robinson elroy!david@csvax.caltech.edu ARPA
david@elroy.jpl.nasa.gov ARPA
{cit-vax,ames}!elroy!david UUCP
Disclaimer: No one listens to me anyway!henry@utzoo.uucp (Henry Spencer) (07/13/89)
In article <1989Jul9.223927.29156@elroy.jpl.nasa.gov> david@elroy.jpl.nasa.gov (David Robinson) writes: >... tweek it a bit for added performence. The first and biggest is >to comment out the c7decode lines if you never expect to get encoded >news... At some point conf/build will probably be modified slightly to at least suggest doing that. The c7decode stuff is kind of specialized. >The other thing I have done which helps if you get many small >batches (ie from a once a minute NNTP sender) is to rewrite 'spacefor' >and 'sizeof' into C if you can. Both are trivial to write but you >may need a statfs() system call for the former... I'm contemplating doing some of this myself, in fact, although if I do use C for spacefor, it will probably popen("df") to do the actual space finding -- that job is just too system-dependent. -- $10 million equals 18 PM | Henry Spencer at U of Toronto Zoology (Pentagon-Minutes). -Tom Neff | uunet!attcan!utzoo!henry henry@zoo.toronto.edu
pim@ctisbv.UUCP (Pim Zandbergen) (07/14/89)
In article <1989Jul12.210254.8631@utzoo.uucp> henry@utzoo.uucp (Henry Spencer) writes: [ about rewriting spacefor in C ] >I'm contemplating doing some of this myself, in fact, although if I do >use C for spacefor, it will probably popen("df") to do the actual space >finding -- that job is just too system-dependent. The following works for me on AT&T System V/386 r 3.2, System V r 3.1.1 on the AT&T 3B2, SCO Xenix, Microport System V/AT, AIX for the IBM PC RT. _______________________________________________________________________________ #include <sys/types.h> #include <sys/stat.h> #include <ustat.h> main() { struct stat statbuf; struct ustat ustatbuf; if (stat("/usr/spool/news", &statbuf) < 0) { perror("cannot fstat"); exit(1); } if (ustat(statbuf.st_dev, &ustatbuf) < 0) { perror("cannot ustat"); exit(1); } printf("%ld free blocks\n", (long)(ustatbuf.f_tfree)); printf("%ld free inodes\n", (long)(ustatbuf.f_tinode)); } _______________________________________________________________________________ It should be noted, however, that this program reports the real number of free blocks, which is not necessarely in 512 byte units. I know of no system call that reports the blocksize of a filesystem *and* is present in all the mentioned flavours of System V (no statfs(2) in AIX nor in uPort V/AT). -- --------------------+----------------------+----------------------------------- Pim Zandbergen | phone: +31 70 542302 | CTI Software BV pim@ctisbv.UUCP | fax : +31 70 512837 | Laan Copes van Cattenburch 70 ...!uunet!mcvax!hp4nl!ctisbv!pim | 2585 GD The Hague, The Netherlands
yun@cinnet.COM (Yun-seng Chao) (07/14/89)
From article <1989Jul12.210254.8631@utzoo.uucp>, by henry@utzoo.uucp (Henry Spencer): > In article <1989Jul9.223927.29156@elroy.jpl.nasa.gov> david@elroy.jpl.nasa.gov (David Robinson) writes: >>The other thing I have done which helps if you get many small >>batches (ie from a once a minute NNTP sender) is to rewrite 'spacefor' >>and 'sizeof' into C if you can. Both are trivial to write but you >>may need a statfs() system call for the former... > > I'm contemplating doing some of this myself, in fact, although if I do > use C for spacefor, it will probably popen("df") to do the actual space > finding -- that job is just too system-dependent. Uh, wasn't there something called "bdf" posted last year that was able to parse the Sys V "df" command? Why not use some of that code for usg systems to do some of this parsing (with acknowledgements to the original author, of course)? -- ------------------------------------------------------------ Yun-seng Chao Cincinnati Network cinnet!yun Voice # (513) 777-5614
henry@utzoo.uucp (Henry Spencer) (07/14/89)
In article <781@ctisbv.UUCP> pim@ctisbv.UUCP (Pim Zandbergen) writes: >>... it will probably popen("df") to do the actual space >>finding -- that job is just too system-dependent. > >The following works for me on >AT&T System V/386 r 3.2, >System V r 3.1.1 on the AT&T 3B2, >SCO Xenix, >Microport System V/AT, >AIX for the IBM PC RT. All of them more or less System V based. That's not all there is to the Unix world, I'm afraid. Of particular note are the systems that are old enough that they have *no* system call to ask about free space (df paws through the raw filesystems). -- $10 million equals 18 PM | Henry Spencer at U of Toronto Zoology (Pentagon-Minutes). -Tom Neff | uunet!attcan!utzoo!henry henry@zoo.toronto.edu
jmm@ecijmm.UUCP (John Macdonald) (07/18/89)
In article <1018@cniysc.cinnet.COM> yun@cinnet.COM (Yun-seng Chao) writes: | |Uh, wasn't there something called "bdf" posted last year that was able to |parse the Sys V "df" command? Why not use some of that code for usg |systems to do some of this parsing (with acknowledgements to the original |author, of course)? "bdf" was a perl program that converted Sys V df output into a format similar to Berkely df. I kind of doubt that Henry and Geoff would want to include perl as a pre-requisite for C news. -- John Macdonald
pjh@mccc.UUCP (Pete Holsberg) (07/21/89)
In article <297@ecijmm.UUCP> jmm@ecijmm.UUCP (John Macdonald) writes: =In article <1018@cniysc.cinnet.COM> yun@cinnet.COM (Yun-seng Chao) writes: =| =|Uh, wasn't there something called "bdf" posted last year that was able to =|parse the Sys V "df" command? Why not use some of that code for usg =|systems to do some of this parsing (with acknowledgements to the original =|author, of course)? = ="bdf" was a perl program that converted Sys V df output into a format =similar to Berkely df. I kind of doubt that Henry and Geoff would want =to include perl as a pre-requisite for C news. =-- =John Macdonald I just cat'd the bdf I have and it's a binary file. I guess there must be two bdf's. -- Pete Holsberg -- Mercer College -- Trenton, NJ 08690 ...!rutgers!njin!princeton!njsmu!mccc!pjh