Cody Boone Ferguson
ffdb219ceb
This was done by:
sgit -e 's/alt code/alternate code/g' . ; git co tmp/OLD_manifest
as well as updating the tmp/manifest.numbers spreadsheet and exporting
to the CSV file (though that would not strictly be necessary).
The way this works means that the update manifest script does not need
to be run as it sgit(1) touched the .entry.json files (any that needed
to be updated) and so would any index.html files. An example for both
.entry.json and index.html files being modified due to manifest is
2012/kang.
Numerous Makefiles were updated, any try scripts that had 'alt code'
instead of 'alternate code' and README.md/index.html files (unrelated to
the manifest as 'alt code' was used in other places too) and obviously
the manifest.numbers spreadsheet.
However to be safe I will do a run of the manifest script in a bit even
though it should not be necessary. If no commits are made it means all
is good and if any commits are made for the manifest it means that I
did not think of something (or else I made another change after this).
To build:
make
Bugs and (Mis)features:
The current status of this entry is:
STATUS: doesn't work with some platforms - please help us fix
For more detailed information see 2011/richards in bugs.html.
To use:
echo expression | ./richards
Try:
./try.sh
Judges' remarks:
Oh no, yet another
dc-style calculator? (see
2000/dlowe using an embedded Perl interpreter and deemed Worst Abuse of the Rules). Well, this one does it (expression calculation, that is,
not the rule abuse) in a different way. The author claims that his calculator
does just-in-time compilation. What would you do to prove or disprove his claim
without attempting to unravel the source?
Author's remarks:
This is an implementation of the classic UNIX
dc command, the
reverse Polish
calculator. Actually it's a pretty bad implementation, since it only implements
32-bit integers (well, sizeof(int)-sized integers). It only implements the
commands +, -, *, /, p (only for numbers), register stacks, macros and
comparisons. Wow, you'd think I could've done better in 2K of code.
OH! I forgot to mention! It's a JIT.
Wait, don't stop reading! I know what you're thinking, platform-specific code is discouraged in IOCCC. But do you see any platform-specific code? There's no assembler here! This JIT works on a dozen platforms, and the only platform-specific assumptions it makes are that the stack grows down and integers are at least 32-bit.
You will probably get warnings while compiling the code. In particular:
-
If your system requires the inclusion of
alloca.hto use alloca, use the compile flag-DAH. -
If your system cannot use
mmap(), use the compile flag-DNM. -
I don't include any headers I don't strictly need to, so you'll get warnings about (at least)
memcpy(),isdigit(),atoi(),calloc(),malloc()andalloca()if you didn't use-DAH.- NOTE from judges: we added these includes in 2023.
-
GCC likes to complain about
while (a = b)(assignments as a condition). -
I don't return from
main(), even though it's declared to returnint.
Platforms on which all known/tested optimization levels work with no additional flags:
- gcc 4.6.2 + x86[_64] + GNU/Linux, Mac OS X or Windows
- clang 2.9-16 + x86[_64] + GNU/Linux or Mac OS X
- tcc 0.9.25
- gcc 4.4.5 + alpha, arm, mips[el], powerpc[64], s390x (IBM zSeries)
- gcc 4.1.3 + vax + NetBSD
Platforms which work but not at all optimization levels or needing special flags:
- owcc 1.9 + x86 + Windows: Needs
-DNM -DAH -fno-stack-check - gcc 4.4.5 + sh4: -O0 works, higher optimization levels don't work for unknown reasons.