Re: DU binaries and /sbin/loader (argv[0] problem?) (FIXED! PATCH!)
Attached is a patch that fixes the Digital UNIX loading code so that argv[0]
is properly set. i.e. ps now reveals (for DU 4.0 netscape):
mcelrath 398 0.3 6.4 22768 16528 ? S 23:05 0:01 netscape
mcelrath 399 0.0 1.5 16872 3896 ? S 23:05 0:00 (dns helper)
With no complaints about /sbin/loader. I think that this is "The Right Way"
to do it. Herbert Riedel's patch (posted a few days ago) seemed a bit hasty
in that it left the bprm->loader variable pointing to the executable, not
/sbin/loader, but this new patch is based on his.
Please test this and let me know your mileage (good *or* bad!).
The patch is also available at ftp://draal.physics.wisc.edu/pub/kernel/
On Tue, 20 Oct 1998, Aideen McConville wrote:
> Hi,
>
> I've tried running some 3.2 DU binaries under RH 5.1 (both 2.0.34 and
> 2.0.35 kernels) and don't know what to do about the "/sbin/loader
> being passed as argv[0]" problem.
>
> Communicator (communicator-v406-export.alpha-dec-osf3.2.tar.gz)
> complains about "/sbin/loader: Fatal Error: loader cannot be invoked
> directly" but then runs ok.
>
> XEmacs (as built for DU3.2) gets very confused, thinking that
> /sbin/loader is supposed to be the xemacs binary!
XEmacs gives me symbol versioning problems (xemacs for DU3.2, I have libs
for DU4.0...) on libcurses.so. Netscape 4.5 complains of unresolved symbols
in the binary!!! (I don't think either of these problems are related to the
/sbin/loader stuff...)
-- Bob
./'^`\._./'~`\._./'~`\._./'~`\._./'#`\._./'~`\._./'~`\._./'~`\._./'^`\.
/ Rube Goldberg? Bah! Amateur! I give you..Windows 98! Linux everywhere \
| Linux, because everyone's work is mission critical. ANYTHING inside. |
|_Bob McElrath (mcelrath@wisconsin.cern.ch) Univ. of Wisconsin at Madison_|
\.___________________________________________________________________./
Reply to: