work in progress

[andrewcmyers]

Here is what I've changed to try to get this working on the Mac on a more recent gcc, though I get the same problems on Linux. Note that I added an autoconf script.

[npe9]

My intuition is that there's a set of incompatible #defines that are leading to it hitting unreachable code. I'll give it a whirl.

[npe9]

I might be short a file. config.status: code/include/acconfig.h is unchanged Configuring garbage collector... ./configure: line 4439: configure: command not found

[andrewcmyers]

Try again?

[npe9]

@andrewcmyers I had a chance to look at the bug on my flight. In this case, "failure: not possible" is coming from a precompiled clu module used for bootstrapping the compiler (_workdir.c). There's something going wrong in there, but since the output is an intermediate representation and not human readable, I need to go through the original source code (_workdir.clu) to see what's going on. I haven't dug too deeply, but it looks like it's trying to parse /etc/passwd and I'd speculate that's a likely point of failure.

[npe9]

It may be silly actually. pclu looks like it requires a "CLU" user. What's not possible here is the CLU user not existing. Let me try and see what happens.

[npe9]

Update, you don't need a CLU user, just $CLUHOME set. That gets it farther:

* thread #1, queue = 'com.apple.main-thread', stop reason = breakpoint 1.1
  * frame #0: 0x000000010017ac20 pclu`_pclu_erstr
    frame #1: 0x000000010010d0b1 pclu`p_envOPscanline(e=CLUREF @ 0x00007ff7bfefdf10) at penv.c:3380:24
    frame #2: 0x000000010010ab15 pclu`p_envOPnext_token(e=CLUREF @ 0x00007ff7bfefdfd0, ret_1=0x00007ff7bfefe018) at penv.c:1764:15
    frame #3: 0x000000010010a2c1 pclu`p_envOPreset(e=CLUREF @ 0x00007ff7bfefe0d0, comp=CLUREF @ 0x00007ff7bfefe0c8, inst=CLUREF @ 0x00007ff7bfefe0c0, errst=CLUREF @ 0x00007ff7bfefe0b8) at penv.c:1376:11
    frame #4: 0x000000010013b0e9 pclu`reset_stuff(e=CLUREF @ 0x00007ff7bfefe2e0, fn=CLUREF @ 0x00007ff7bfefe2d8) at top2.c:1219:11
    frame #5: 0x000000010013a0c0 pclu`compile1(e=CLUREF @ 0x00007ff7bfefe4c0, fn=CLUREF @ 0x00007ff7bfefe4b8, ret_1=0x00007ff7bfefe4e8) at top2.c:611:11
    frame #6: 0x0000000100139879 pclu`compile_IB_1(iv_1=CLUREF @ 0x00007ff7bfefe500, locals=0x00007ff7bfefe968, iecode=0x00007ff7bfefe848) at top2.c:250:23
    frame #7: 0x00000001001374ef pclu`all_input_names_IB_1(iv_1=CLUREF @ 0x00007ff7bfefe530, locals=0x00007ff7bfefe830, iecode=0x00007ff7bfefe848) at top1.c:4021:19
    frame #8: 0x0000000100137e75 pclu`all_derived_names(fs=CLUREF @ 0x00007ff7bfefe7f0, suffix=CLUREF @ 0x00007ff7bfefe7e8, tyo=CLUREF @ 0x00007ff7bfefe7e0, proc=(pclu`all_input_names_IB_1 at top1.c:4015), user_locals=0x00007ff7bfefe830, iecode=0x00007ff7bfefe848) at top1.c:4297:22
    frame #9: 0x00000001001372ca pclu`all_input_names(args=CLUREF @ 0x00007ff7bfefe8b0, suffix=CLUREF @ 0x00007ff7bfefe8a8, tyo=CLUREF @ 0x00007ff7bfefe8a0, proc=(pclu`compile_IB_1 at top2.c:243), user_locals=0x00007ff7bfefe968, iecode=0x00007ff7bfefe970) at top1.c:3960:30
    frame #10: 0x00000001001393d0 pclu`compile(mode=CLUREF @ 0x00007ff7bfefea00, args=CLUREF @ 0x00007ff7bfefe9f8, outst=CLUREF @ 0x00007ff7bfefe9f0, ret_1=0x00007ff7bfefea68) at top2.c:86:22
    frame #11: 0x000000010012fc74 pclu`process_commands(line=CLUREF @ 0x00007ff7bfefebf0, tyo=CLUREF @ 0x00007ff7bfefebe8, from=CLUREF @ 0x00007ff7bfefebe0, ret_1=0x00007ff7bfefec38) at top1.c:695:27
    frame #12: 0x00000001000f3092 pclu`initialize_specs at init.c:353:11
    frame #13: 0x00000001000f2c9a pclu`initialize_compiler(ret_1=0x00007ff7bfefef58) at init.c:211:11
    frame #14: 0x000000010012ee3b pclu`start_up at top1.c:98:19
    frame #15: 0x000000010017a984 pclu`main_2 + 388
    frame #16: 0x000000010039151e dyld`start + 462

I believe it's trying to reset the environment and failing to parse. I'll confirm the hypothesis in a little bit.

[andrewcmyers]

I noticed the CLUHOME issue but was not clear on what to set it to

[npe9]

Brief update, the cause of the failed scanline is a failure to recognize the tag (EOF in this case) in stream$getc. It's getting to the tagcase statement in https://github.com/npe9/pclu/blob/master/sys/clu/stream.clu#L373 (which was precompiled to https://github.com/npe9/pclu/blob/master/code/sysclu/Opt/stream.c#L2305).

[npe9]

@andrewcmyers adding that back in fixed it:

gives us

Creating DU specs from /Users/npe/src/pclu/lib/clu/_int_range.clu
time = 0.000
Creating DU specs from /Users/npe/src/pclu/lib/clu/bigint.clu
time = 0.001
Creating DU specs from /Users/npe/src/pclu/lib/clu/dparse.clu
time = 0.000
Creating DU specs from /Users/npe/src/pclu/lib/clu/heap.clu
time = 0.000
Creating DU specs from /Users/npe/src/pclu/lib/clu/lex_scan.clu
time = 0.000
Creating DU specs from /Users/npe/src/pclu/lib/clu/os.clu
time = 0.000
Creating DU specs from /Users/npe/src/pclu/lib/clu/parse_line.clu
time = 0.000
Creating DU specs from /Users/npe/src/pclu/lib/clu/quicksort.clu
time = 0.000
Creating DU specs from /Users/npe/src/pclu/lib/clu/quick_sort.clu
time = 0.000
Creating DU specs from /Users/npe/src/pclu/lib/clu/scase.clu
time = 0.000
Creating DU specs from /Users/npe/src/pclu/lib/clu/split_line.clu
time = 0.000
Creating DU specs from /Users/npe/src/pclu/lib/clu/string_lindexs.clu
time = 0.000
Creating DU specs from /Users/npe/src/pclu/lib/clu/string_ops.clu
time = 0.000
Creating DU specs from /Users/npe/src/pclu/lib/clu/strim.clu
time = 0.000
Creating DU specs from /Users/npe/src/pclu/lib/clu/xref.clu
time = 0.001
Creating DU specs from /Users/npe/src/pclu/lib/random.spc
time = 0.000
Dumped to /Users/npe/src/pclu/misc.lib

Lots of researchware code to fix for sure, but we at least have the compiler pretending to work.

[npe9]

@andrewcmyers next question, what are you looking to use clu for? LARCH?

Edit: Also, I made a mistake and put the previous comment in the non-autotools pull request. The next step (if we need it) is to bootstrap the compiler from itself.

Edit2: I completely forgot I had corralled the larch distro during my initial interest in LARCH/CLU https://github.com/npe9/larch

[andrewcmyers]

Actually I am writing a chapter about CLU for a book and I want to make sure my example code is correct!

[npe9]

@andrewcmyers I'd love to see it when you're done! Thanks for getting me back and interested in CLU. We do a lot of specification work and it's nice to go back in time a bit.