self@BAYES.ARC.NASA.GOV (Matthew Self) (10/01/88)
Hello, Despite following the directions in the INSTALL file carefully, and trying several times, I am unable to get a working gcc on my Sun-3 which runs SunOS 4.0, using either 1.27 or 1.28. The first make (using cc) generates several "statement not reached warnings" (see dribble files below). Since there was no mention of this in the INSTALL file, I suspect that there is already a problem at this point. The second make generates a "warning: floating point number exceeds range of double" for the float constant 0.0 (again see dribble file below). The third pass is identical and does pass the comparison test. The installed gcc generates the same floating point error for any float constant of 0.0, among other errors (which I need to document better before reporting). Does everyone have a problem doing this installation, or am I unlucky? (or careless?) I was careful not to set the FLOAT_OPTION environment variable. I tried both 1.27 and 1.28 and they have the same behaviour. There were no checksum errors in the tar files which I obtained by ftp. The only change I made was to change the execuatbles directory from /usr/local to /usr/local/bin at the beginning of the makefile. I used the sun4 config file, the sun3 machine description, and the 68k md and aux-output files. The only thing which might be unusual about our Sun setup is that we have a hacked libc.so which uses the resolver directly rather than YP. Could this account for the errors I saw? On a more positive note, I wish to commend everyone there on an incredible compiler! (Judging by experience running it on another system....) I couldn't live without variable length arrays! Such a nice simple extension. (Is there *any* chance of ANSI or someone adopting something like it?) Hope I got my question to the right place (sorry if I missed)... Matthew Self self@ptolemy.arc.nasa.gov