Dear Torbjörn, > This is not yet isolated. It looks like a compiler or perhaps more > likely a linker bug. Last night I reverted another change which > benefited gcc202, so now there will be more errors. > > I am not actively working on this issue, but I will probably isolate it > at some point. > > Alas, I got around to this now. > > There were a number of problems: > > 1. The pseudo-insn setx didn't work on gcc202. I worked around this by > expanding the needed insn, and also to go with the 44-bit addressing > model (instead of the expensive 64bit). 44-bit addrs is the default > on both gcc202 and NetBSD 7.x at least. (I suppose we should allow > an configure option for this, but since my SPARC access is extremely > limited I cannot test any such features. Besides, SPARC is not in > much use.) > > 2. The relocs used for PIC in sparc32/sparc-defs.m4 fails on gcc202. I > updated with PICkier choices. > > 3. The C sec_invert and the Sparc T3 asm cnd_add_n/cnd_sub_n did not > agree about how to specify the condition. Should true be 1 or any > non-zero value? I decided for the latter and updated the asm.
thank you, I confirm make check runs with the gmp-6.1.99-20170831 snapshot on gcc202. However I had to define INTERPRETER= in test-driver. Also the "Testsuite summary" is still not very informative: ============================================================================ Testsuite summary for GNU MP 6.1.99 ============================================================================ # TOTAL: 0 # PASS: 0 # SKIP: 0 # XFAIL: 0 # FAIL: 0 # XPASS: 0 # ERROR: 0 Paul _______________________________________________ gmp-bugs mailing list [email protected] https://gmplib.org/mailman/listinfo/gmp-bugs
