Re: Issues with compiler failures

[Paul Koning <paulkoning%comcast.net@localhost>]

> On Mar 24, 2019, at 3:49 PM, Johnny Billquist <bqt%update.uu.se@localhost> wrote:
> 
> ...
> Ran through a new update yesterday, and tried again.
> Got a different problem, but in the end still fails building gcc itself.
> (Which is kindof disappointing, since last I tried before this round, I usually was hitting problems later than this.)
> 
> Problem this time:
> 
> Huff:/usr/src# uname -a
> NetBSD Huff.BQTnet.SE 8.99.36 NetBSD 8.99.36 (Krille) #3: Fri Mar 22 06:30:30 CET 2019 root%GW.SoftJAR.SE@localhost:/usr/obj/sys/arch/vax/compile/Krille vax
> 
> ...
> 
> c++ -fno-PIE -c  -DIN_GCC_FRONTEND -O -DIN_GCC -DCROSS_DIRECTORY_STRUCTURE   -fno-exceptions -fno-rtti -fasynchronous-unwind-tables -W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual -Wmissing-format-attribute -Woverloaded-virtual -pedantic -Wno-long-long -Wno-variadic-macros -Wno-overlength-strings -DHAVE_CONFIG_H -I. -Ic -I/usr/src/tools/gcc/../../external/gpl3/gcc.old/dist/gcc -I/usr/src/tools/gcc/../../external/gpl3/gcc.old/dist/gcc/c -I/usr/src/tools/gcc/../../external/gpl3/gcc.old/dist/gcc/../include -I/usr/src/tools/gcc/../../external/gpl3/gcc.old/dist/gcc/../libcpp/include -I/usr/src/obj/tooldir.NetBSD-8.99.36-vax/include -I/usr/src/obj/tooldir.NetBSD-8.99.36-vax/include -I/usr/src/obj/tooldir.NetBSD-8.99.36-vax/include -I/usr/src/tools/gcc/../../external/gpl3/gcc.old/dist/gcc/../libdecnumber -I/usr/src/tools/gcc/../../external/gpl3/gcc.old/dist/gcc/../libdecnumber/dpd -I../libdecnumber -I/usr/src/tools/gcc/../../external/gpl3/gcc.old/dist/gcc/../libbacktrace  -DNETBSD_TOOLS -DTARGET_SYSTEM_ROOT=0 -DTARGET_SYSTEM_ROOT_RELOCATABLE -o c/c-decl.o -MT c/c-decl.o -MMD -MP -MF c/.deps/c-decl.TPo /usr/src/tools/gcc/../../external/gpl3/gcc.old/dist/gcc/c/c-decl.c
> /usr/src/tools/gcc/../../external/gpl3/gcc.old/dist/gcc/c/c-decl.c: In function 'tree_node* finish_enum(tree, tree, tree)':
> /usr/src/tools/gcc/../../external/gpl3/gcc.old/dist/gcc/c/c-decl.c:8193:1: error: unrecognizable insn:
> }
> ^
> (insn 72 71 73 16 (set (reg:SI 165)
>        (subreg:SI (mem:DI (plus:SI (mult:SI (reg:SI 162)
>                        (const_int 8 [0x8]))
>                    (reg/f:SI 103 [ _188 ])) [14 *_206+0 S8 A32]) 4)) /usr/src/tools/gcc/../../external/gpl3/gcc.old/dist/gcc/wide-int.h:795 -1
>     (nil))
> /usr/src/tools/gcc/../../external/gpl3/gcc.old/dist/gcc/c/c-decl.c:8193:1: internal compiler error: in extract_insn, at recog.c:2287
> no stack trace because unwind library not available
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <http://www.NetBSD.org/support/send-pr.html> for instructions.
> nbgmake[1]: *** [c/c-decl.o] Error 1
> nbgmake[1]: Leaving directory `/usr/obj/tools/gcc/build/gcc'
> nbgmake: *** [all-gcc] Error 2
> 
> *** Failed target:  .build_done

It looks like it's doing something like this:

	int64_t x[100];
	int32_t foo = x[bar];

and it's trying to generate a CVTQL which unfortunately does not exist.  So it looks like there's something in the machine description that tells GCC that it is allowed to try to do this, but then later when it wants to generate the code for that, it fails because there actually isn't any such instruction.

	paul