build fails with ../../snek-gram.ll:258:5: error: ‘for_depth’ undeclared

[fanoush]

I am trying to build from source on debian 11 bullseye i386 (old laptop) and aarch64 (raspberry Pi4 with 64bit OS) and on both of them it fails like this

make[1]: Entering directory '/home/pi/snek/ports/posix'
lola -DSNEK_SLICE -DSNEK_DICT -o snek-gram.h ../../snek-gram.ll
python3 ../../snek-builtin.py ../../snek-keyword.builtin ../../snek-base.builtin snek-posix.builtin ../../snek-math.builtin ../../snek-input.builtin -o snek-builtin.h
  CC  snek-builtin.o
  CC  snek-code.o
  CC  snek-exec.o
  CC  snek-error.o
  CC  snek-frame.o
  CC  snek-func.o
  CC  snek-lex.o
  CC  snek-list.o
  CC  snek-memory.o
  CC  snek-name.o
  CC  snek-parse.o
In file included from ../../snek-parse.c:219:
../../snek-gram.ll: In function ‘parse’:
../../snek-gram.ll:258:5: error: ‘for_depth’ undeclared (first use in this function)
  258 |     for_depth--;
      |     ^~~~~~~~~
../../snek-gram.ll:258:5: note: each undeclared identifier is reported only once for each function it appears in
In file included from ../../snek-parse.c:219:
../../snek-gram.ll:17:11: error: ‘snek_print_val’ undeclared (first use in this function); did you mean ‘snek_print’?
   17 | command  : @{ snek_print_val = snek_interactive; }@ stat
      |           ^~~~~~~~~~~~~~
      |           snek_print
make[1]: *** [../../snek.defs:163: snek-parse.o] Error 1
make[1]: Leaving directory '/home/pi/snek/ports/posix'
make: *** [Makefile:30: ports/posix/snek] Error 2

I can see that mising stuff is declared in C macro PARSE_TOP here https://github.com/keith-packard/snek/blob/main/snek-parse.c#L39 but I don't see the macro being used anywhere so I'm not sure how this is supposed to work. also in generated ports/posix/snek-gram.h macro PARSE_TOP is not used anywhere.

I installed dependencies including lola as per readme, lola was downloaded as

http://deb.debian.org/debian bullseye/main arm64 lola all 1.4-1 [12.0 kB]   

I am trying default main branch if that matters.

[fanoush]

Oh, well, I guess the answer is here https://github.com/keith-packard/lola/commit/0b09e90d17664e5f7391f5b749cfeed1cb488530 Looks like lola version 1.4 in latest debian is too old (?) so the README.md is too optimistic :-(

[fanoush]

OK sudo vi $(which lola) and pasting those three lines there fixed the build.

BTW at first I was guessing gawk and some gnu extension since after second read of readme you mention gawk but it is not in install instructions (and bullseye does not install gawk by default, there is mawk) but that was not it, mawk is probably good enough.

[keith-packard]

Oh, well, I guess the answer is here keith-packard/lola@0b09e90 Looks like lola version 1.4 in latest debian is too old (?) so the README.md is too optimistic :-(

I couldn't find anything in README.md that even mentions the version of lola required; it only talks about installing lola from debian unstable. That's not wrong, but it's also not very helpful.

[keith-packard]

OK sudo vi $(which lola) and pasting those three lines there fixed the build.

BTW at first I was guessing gawk and some gnu extension since after second read of readme you mention gawk but it is not in install instructions (and bullseye does not install gawk by default, there is mawk) but that was not it, mawk is probably good enough.

Not having gawk won't stop the programs from being built, but it will prevent the build from completing successfully as the snek build scripts use a GNU extension, strtonum, which isn't available in mawk.

[fanoush]

That's not wrong, but it's also not very helpful.

Yes, indeed. It is I who was too optimistic :-) Referring to stable and unstable is not very specific so I was just hoping it was written when bullseye was not released yet. But that was bad guess as most probably unstable -> bullseye switch was done much earlier than its release date.

As for gawk - after first failure I actually retried only inside ports/posix so didn't notice mawk failure. Anyway putting gawk into that # apt install lola gcc-avr .... line in README could help too.

I'll leave this open as merging the PR will close it but thanks, problem solved.