OK, I must have been doing something wrong. I ran "make distclean" for everything I have built up to this point and tried again. the problem went away.
Sorry for the false alarm. Tom On Mon, Nov 11, 2013 at 1:14 PM, Tom Kacvinsky <[email protected] > wrote: > It helps if I tell you I am using opnessl-1.0.1e > > > > On Mon, Nov 11, 2013 at 1:06 PM, Tom Kacvinsky < > [email protected]> wrote: > >> The short story is that I want to make static libraries of Open SSL so >> that I can make Python modules (_ssl) that link against Open SSL >> statically, thereby reducing the amount of run time dependencies we have. >> But when I link these libraries in statically, I get relocation errors form >> the linker due to some of the code not being PIC. >> >> I don't know if the development is the appropriate forum for this, but I >> think it might since it has to do with how Open SSL builds on Solaris. >> >> Here is my set up: gcc 4.7.4 with GNU assembler and Solaris linker (gcc >> was configured with --with-gnu-as, but not with --with-gnu-ld). I ran >> >> ./config --prefix=/home/tjk/openssl no-shared -fPIC >> >> The configured build is for 32 bit Sparc Solaris running on a sparcv9 >> processor. >> >> The C code compiles into PIC objects just fine, but the GNU assembler >> does not produce PIC code for sparccpuid.S. I worked around this by first >> preprocessing the file using gcc: >> >> gcc -I... -D... -E sparcpuid.S > sparcpuid.s >> >> and then assembled it using >> >> as -V -Qy -s -xarch=v8plusa -K PIC -xcode=pic32 sparcpuid.s >> >> (I first ran gcc -v to see what gcc was going to pass to the assembler, >> then took that output and modified it so that the right options would be >> used for the Solaris assembler). I found that the crucial option is >> -xcode=pic32. >> >> I do not really ant to do this as I plan on scripting builds of Open SSL >> and this requires manual intervention. Do you know of a way to get the GNU >> assemble to produce PIC code? >> >> Thanks, >> >> Tom >> >> >> >
