[2152] in Kerberos_V5_Development
Re: thoughts about shared library build process
daemon@ATHENA.MIT.EDU (Theodore Y. Ts'o)
Tue Dec 31 04:04:01 1996
Date: Tue, 31 Dec 1996 04:03:49 -0500
From: "Theodore Y. Ts'o" <tytso@MIT.EDU>
To: Tom Yu <tlyu@MIT.EDU>
Cc: "Theodore Y. Ts'o" <tytso@MIT.EDU>, Tom Yu <tlyu@MIT.EDU>, krbdev@MIT.EDU
In-Reply-To: Tom Yu's message of Sun, 29 Dec 1996 19:31:44 -0500,
<9612300031.AA10260@tesla-coil.MIT.EDU>
Date: Sun, 29 Dec 1996 19:31:44 -0500
From: Tom Yu <tlyu@MIT.EDU>
The issue of pulling in everything at once with dlopen is going to be
kinda tricky if we're pulling in libkdb and it needs symbols from
libdb... maybe we should bite the bullet and hack up libdb so that it
can be built shared. We could even set it up so that it checks to see
if it's being built in the krb5 tree before pulling in the krb5
macros.
Pulling in libkrb5.so and libgssapi.so via dlopen() is **far** more
important than being able to pull in libkdb.so. Building libdb shared
is definitely an interesting project, and now that Marc is willing to
make changes to util/db2 (such as renaming all of the conflicting
types), we could consinder this. But I hardly consider pulling in
libkdb from dlopen() a compelling reason by itself to say that we should
start building libdb shared.
- Ted
