Solved: Problem rebuilding Ray's lz77 utility with PureC

[lp]

I contacted Ray, but he could not remember exactly how he compiled the file lz77.s (v1.3) in the archive. He thought he used an old gcc with the -02 option. However, it appears to be compiled PureC with a modified startup module after looking it over in a dis-assembler. It builds without any warnings or errors, but when I compress the same test file the resulting file sizes don't match the output of his original utility. Any PureC experts here have some ideas?

http://s390174849.online.de/ray.tscc.de/code.htm#lz77

It also dumps Warning: use TCFLTLIB to get float support for printf to the console when running. Can't figure out why.

[shoggoth]

I contacted Ray, but he could not remember exactly how he compiled the file lz77.s (v1.3) in the archive. He thought he used an old gcc with the -02 option. However, it appears to be compiled PureC with a modified startup module after looking it over in a dis-assembler. It builds without any warnings or errors, but when I compress the same test file the resulting file sizes don't match the output of his original utility. Any PureC experts here have some ideas?

http://s390174849.online.de/ray.tscc.de/code.htm#lz77

It also dumps Warning: use TCFLTLIB to get float support for printf to the console when running. Can't figure out why.

Different integer size?

(use stdint.h, everyone!)

[lp]

Not sure how stdint.h should help me since I'm using the correct compiler (PureC).

[shoggoth]

Not sure how stdint.h should help me since I'm using the correct compiler (PureC).

My bad. I was sure I built it w. GCC and it worked right away (hence my suspicion about integer size...).

[lp]

Ray's binary contains symbols and if I enable symbols in PureC they are the same when compared to my build. It just don't work right. lol Did you do actual compression tests to insure your build generates the same output as his?

There is one odd thing. The title message in the binary he supplied doesn't match the code in the source file. Possibly he included an older source lising by mistake.

[lp]

I just tried the gcc cross-tools and adding the include you suggested. The binary ends up about 90kb larger and the same issue. Compression results do not match Rays' build, but matches my PureC build. Somehow his build manages to compress 3kb smaller. I think I can safely say he didn't actually include the correct source file in the archive or he altered the dictionary size or some other compression aspect.

[lp]

The readme states gwen found the final bug that led to v1.3. I finally found the correct combination of PureC options to produce a binary match, at least until I reached the encoder. Ray sent me a link to v1.2 in his reply so I ran a diff on the sources. One line of code in the encoder was changed. On a whim I pasted the v1.2 line into the v1.3 source and tada, now my compression tests are identical. The question remains, which one is the bug fixed version, the source listing (compresses less) or the binary (compresses more). lol

One problem remains, why this line dumps the odd floating point warning. The source compiles fine and doesn't complain about anything.
printf("ok.\nAchieved ratio: %.2f%%\n", (float)codesize*100 / textsize);

Something I need to include? I have these libraries in the project file, there is no TCFLTLIB included with PureC as far as I can tell.
;libs
PCSTDLIB.LIB ;standard library
PCFLTLIB.LIB

[lp]

Switched the library order in the project file. Problem solved.
;libs
PCFLTLIB.LIB
PCSTDLIB.LIB ;standard library

[wongck]

libraries gotchas are always difficult to see at first.

[lp]

Combine that with GFA user trying to use 'C'.

[wongck]

but the programming logic is the same, just keywords and sentences construct may be different.
well, one single typo in C can generate 100 errors and warnings.