andrewcmyers/master

[npe9]

• Add missing prototypes.
• Check in config file for now. Should be autogenerated.
• Simple autoconf script.
• The actual configuration script
• Fix prototype
• Fix newlines
• Make PCLU driver program build again
• Don't assume . in path
• initial autotools support to make it easier to debug the compiler

[npe9]

@andrewcmyers I got sick of futzing with the makefiles so I made an initial autotools setup to get the compiler building with common compilers and flags. Make sure I haven't messed up too much.

This new setup allowed me to dig deeper into the problem. The relevant backtrace is:

* thread #1, queue = 'com.apple.main-thread', stop reason = step in
  * frame #0: 0x00000001000f2410 pclu`_chanOPgetb(chref=CLUREF @ rbx, bv=CLUREF @ rsi, ans=0x00007ff7bfefe660) at _chan.c:1381:18 [opt]
    frame #1: 0x00000001000bf7e6 pclu`streamOPgetc(st=CLUREF @ r12, ret_1=0x00007ff7bfefe6d0) at stream.c:2334:23 [opt]
    frame #2: 0x0000000100094dce pclu`p_envOPscanline(e=<unavailable>) at penv.c:3020:19 [opt]
    frame #3: 0x00000001000942c3 pclu`p_envOPnext_token(e=CLUREF @ rbx, ret_1=0x00007ff7bfefe768) at penv.c:1764:15 [opt]
    frame #4: 0x0000000100093df3 pclu`p_envOPreset(e=CLUREF @ rbx, comp=CLUREF @ r14, inst=CLUREF @ r15, errst=CLUREF @ r12) at penv.c:1376:11 [opt]
    frame #5: 0x00000001000b103d pclu`reset_stuff(e=CLUREF @ rbx, fn=CLUREF @ r13) at top2.c:1219:11 [opt]
    frame #6: 0x00000001000b0859 pclu`compile1(e=CLUREF @ r13, fn=CLUREF @ r14, ret_1=0x00007ff7bfefe898) at top2.c:611:11 [opt]
    frame #7: 0x00000001000b0475 pclu`compile_IB_1(iv_1=<unavailable>, locals=0x00007ff7bfefea60, iecode=0x00007ff7bfefe9d8) at top2.c:250:23 [opt]
    frame #8: 0x00000001000aeda9 pclu`all_input_names_IB_1(iv_1=<unavailable>, locals=0x00007ff7bfefe9c0, iecode=0x00007ff7bfefe9d8) at top1.c:4021:19 [opt]
    frame #9: 0x00000001000af3d9 pclu`all_derived_names(fs=CLUREF @ rbx, suffix=<unavailable>, tyo=<unavailable>, proc=<unavailable>, user_locals=<unavailable>, iecode=0x00007ff7bfefe9d8) at top1.c:4297:22 [opt]
    frame #10: 0x00000001000aeb4e pclu`all_input_names(args=CLUREF @ rbx, suffix=<unavailable>, tyo=<unavailable>, proc=<unavailable>, user_locals=<unavailable>, iecode=0x00007ff7bfefea68) at top1.c:3960:30 [opt]
    frame #11: 0x00000001000b01c9 pclu`compile(mode=<unavailable>, args=<unavailable>, outst=CLUREF @ r15, ret_1=0x00007ff7bfefeb70) at top2.c:86:22 [opt]
    frame #12: 0x00000001000aa227 pclu`process_commands(line=<unavailable>, tyo=CLUREF @ 0x00007ff7bfefeb60, from=<unavailable>, ret_1=<unavailable>) at top1.c:695:27 [opt]
    frame #13: 0x0000000100086833 pclu`initialize_specs at init.c:353:11 [opt]
    frame #14: 0x00000001000865a3 pclu`initialize_compiler(ret_1=0x00007ff7bfefec80) at init.c:211:11 [opt]
    frame #15: 0x00000001000a9b40 pclu`start_up at top1.c:98:19 [opt]
    frame #16: 0x000000010010200b pclu`main_2(_argc=<unavailable>, _argv=<unavailable>, _envp=<unavailable>) at util.c:233:11 [opt]
    frame #17: 0x000000010032951e dyld`start + 462

It looks like the token stream is invalid (it's bombing out on _chan$getb because the stream reference is poisoned). This is a huge PITA to debug because the compiler code is intermediate representation from a bootstrapped clu compiler. Cross referencing with the original clu files helps.

[npe9]

Looks like the poison is here: https://github.com/npe9/pclu/blob/master/code/sysasm/Opt/_chan.c#L170

So the next hypothesis is that the mode isn't getting set correctly when the channel is being opened.

Edit: a quick trip through the debugger confirms it.

[andrewcmyers]

I may have changed line 172 because it was clearly bogus but maybe it was compensating for an error elsewhere. That !flags&1 is bad code but i think it causes rd to be set when the mode is read only. Note that ! has higher precedence than &

[andrewcmyers]

The logic of this code is just wrong. You can't write stuff like (flags & O_RDWR). You might write (flags & O_ACCMODE) == O_RDWR, but the flags don't work the way this code seems to think they do.