rfg@nsc.nsc.com (Ron Guilmette) (11/25/88)
In a recent posting to bug-gcc, sun!apple.com!satyr!local writes: >My system is: Altos 1000 running Unix System 5.3.1b. It's a >386 box. > >I made the stage1 and stage2 compilers. Everything was fine. >Running the compare on the object files, I always get: > >c-convert.o stage2/c-convert.o differ: char 5, line 1 >version.o stage2/version.o differ: char 5, line 1 >.... >ad nauseam... > >That is, every object file differs at char 5, line 1. I have *always* experienced the same thing when building GCC on System V. I learned long ago to just ignore it. I have always figured that as long as I could successfully get through building all 3 stages then things must be mostly right. >Examining ho.o and ho.o.orig, the *only* place they differ is character >5, which seems to be set to a random value at each compilation (with the >same compiler!). > >Any ideas??? It seems that perhaps some static memory is not getting >correctly initialized... I doubt that the differences are due to "random" values. My theory is that "date/time of compilation" information is being put into the COFF object files so that they will always be different for different compilation dates/times. I have no evidence for this view. It is just a theory based on total ignorance. Ron Guilmette National SemiConductor, 1135 Kern Ave. M/S 7C-266; Sunnyvale, CA 94086 Internet: rfg@nsc.nsc.com or amdahl!nsc!rfg@ames.arc.nasa.gov Uucp: ...{pyramid,sun,amdahl,apple}!nsc!rfg