OK, I have what I believe are a set of "good" patches which "cleanly" port
nessus to the amd64 64 bit application environment. I will be uploading
these patches later today.
Some comments --
1. For casting types, I am using mostly "ssize_t". Microsoft recommends
using size_t (and ssize_t) since they are polymorphic on their Windows
systems (4 byte on 32 bit systems and 8 bytes on 64 bit systems ... same as
pointers). I could have used "long" in many places but there appears to be
more code intended to run under MS Windows than I originally realized. For
consistency, I am using "ssize_t" (or "size_t" where that is appropriate).
2. There are some patches which add function prototypes. All of these
produced cast problems on the amd64 and do not appear to cause problems on
the i386, etc. I have mentioned memmem() before. Without the prototype,
there are cast warnings. Even if a system does not have memmem(), then the
nessus "slow" one defined in strutils.c is used ... and a prototype is still
needed.
3. I found a library setup problem with libtool. I was getting strange
warnings about libraries not shared from libtool on the amd64. I found that
nessus-config had "$PREFIX/lib" hard coded. The patch changes that to
"@libdir@" which becomes "/usr/lib" on the i386 and "/usr/lib64" on the
amd64. So, no more strange messages from libtool on the amd64. However, now
I get some strange (different) messages from libtool on the i386 .. can't
win.
4. I ran the compiles with "-Wall -pedantic" to try and get things sorted out.
Currently, everything compiles clean on the amd64 except for
nessus-core/nessus/regex.c ... I get some warnings about print format
discrepancies. On the i386, nessus-core/nessus/regex.c compiles clean but
some other files now get warnings about "long int format, int arg" [and I am
using a ssize_t cast].
I welcome comments, corrections, etc. Hopefully, these patches are "system
independent" (runs on *BSD, Solaris, IRIX, etc.). If not, please point out
my errors.
--
Gene Czarcinski
nessus to the amd64 64 bit application environment. I will be uploading
these patches later today.
Some comments --
1. For casting types, I am using mostly "ssize_t". Microsoft recommends
using size_t (and ssize_t) since they are polymorphic on their Windows
systems (4 byte on 32 bit systems and 8 bytes on 64 bit systems ... same as
pointers). I could have used "long" in many places but there appears to be
more code intended to run under MS Windows than I originally realized. For
consistency, I am using "ssize_t" (or "size_t" where that is appropriate).
2. There are some patches which add function prototypes. All of these
produced cast problems on the amd64 and do not appear to cause problems on
the i386, etc. I have mentioned memmem() before. Without the prototype,
there are cast warnings. Even if a system does not have memmem(), then the
nessus "slow" one defined in strutils.c is used ... and a prototype is still
needed.
3. I found a library setup problem with libtool. I was getting strange
warnings about libraries not shared from libtool on the amd64. I found that
nessus-config had "$PREFIX/lib" hard coded. The patch changes that to
"@libdir@" which becomes "/usr/lib" on the i386 and "/usr/lib64" on the
amd64. So, no more strange messages from libtool on the amd64. However, now
I get some strange (different) messages from libtool on the i386 .. can't
win.
4. I ran the compiles with "-Wall -pedantic" to try and get things sorted out.
Currently, everything compiles clean on the amd64 except for
nessus-core/nessus/regex.c ... I get some warnings about print format
discrepancies. On the i386, nessus-core/nessus/regex.c compiles clean but
some other files now get warnings about "long int format, int arg" [and I am
using a ssize_t cast].
I welcome comments, corrections, etc. Hopefully, these patches are "system
independent" (runs on *BSD, Solaris, IRIX, etc.). If not, please point out
my errors.
--
Gene Czarcinski