emv@math.lsa.umich.edu (Edward Vielmetti) (04/10/90)
os: sunos 4.0.3, running with the "sun-fixes" version of libc_resolv.so that puts name-server based resolver routines in the SunOS shared library. This is nice as far as it goes, but since Sun released these shared libraries last July there's been some good ideas and useful improvements in resolver implementations. The ones I'd like to see inserted into my environment are as follows: DOMAINPATH -- an environment variable which controls the search list so that "telnet ummts" would try "ummts.math.lsa.umich.edu", "ummts.stat.lsa.umich.edu", "ummts.cc.umich.edu" in that order. res_search -- fix to not append the local domain if there are at least two dots in the name, to avoid unneeded queries on fully specified names, i.e. none of this zip.eecs.umich.edu.math.lsa.umich.edu zip.eecs.umich.edu.lsa.umich.edu zip.eecs.umich.edu.umich.edu lookup nonsense. hesoid stuff -- support for HS class and TXT records. This is all within the constraint that I don't want to give up shared libraries! I have these fixes, now, I'm just wondering, what it takes to make them fit into the shared library regime. Sun apparently provides the entire libc.a in a position independent (pic) format, so it's reasonable to think of inserting in the updated .o's if they could be generated. Thanks for any help. x-posted to comp.mail.mh cause I know there's some shared library expertise there; followup to the domains list. --Ed Edward Vielmetti, U of Michigan math dept. emv@math.lsa.umich.edu
emv@math.lsa.umich.edu (Edward Vielmetti) (04/13/90)
In article <EMV.90Apr9205846@duby.math.lsa.umich.edu> emv@math.lsa.umich.edu (Edward Vielmetti) writes:
os: sunos 4.0.3, running with the "sun-fixes" version of libc_resolv.so
that puts name-server based resolver routines in the SunOS shared library.
This is all within the constraint that I don't want to give up shared
libraries! I have these fixes, now, I'm just wondering, what it takes
to make them fit into the shared library regime. Sun apparently
provides the entire libc.a in a position independent (pic) format, so
it's reasonable to think of inserting in the updated .o's if they
could be generated.
fixed in 4.1 apparently -- the resolv.a pieces build clean with
`cc -pic' (tested with `ld -assert pure-text *.o') and I am assured
that all the stuff you need to build a new libc.so from the old
one is there.
sigh, if all that stuff is "fixed in 4.1", I can hardly wait to
hear what's "broken in 4.1"....
--Ed
Edward Vielmetti, U of Michigan math dept.
emv@math.lsa.umich.edu