Fixes to thanks for fixes file

At least one entry was missing (1984/anonymous).

Some entries were removed. One such entry was not ever broken but
instead when I had renamed a file (to markdown) I forgot to update the
DATA variable in the Makefile which then broke compilation. This was
seen in the log and mistakenly believed that there was something wrong
with the entry when there wasn't. Other entries removed were not fixes
but notes.

Various README.md files have been updated for this sometimes just
formatting merely because I opened the file because for instance there
is an alternate version (many by me) that is referred to. I'd not be
surprised if more README.md files have to be updated for this much
better system but these will be found in time when I do another pass
(and some years I have not checked at all yet).
This commit is contained in:
Cody Boone Ferguson committed 2023-07-01 09:49:26 -07:00
1 parent a7f9a722fc
commit 76cef25333
11 files changed
+405 -367

No files matched your search

+12 -16
View File
@@ -64,12 +64,12 @@ other types of files?
### Alternate code:
The alternate code allows this entry to work successfully even with clang. [Cody
Boone Ferguson](/winners.html#Cody_Boone_Ferguson) notes that unfortunately it's
something of a hack or maybe even a kludge. This however seems important because
it works well and allows for not messing with the `s` string which is quite
complicated. Even changing the length could break functionality and indeed the
length would have to be longer for this to work with clang.
The alternate code allows this entry to work successfully even with clang.
Unfortunately it's something of a hack or maybe even a kludge. This however
seems important because it works well and allows for not messing with the `s`
string which is quite complicated. Even changing the length could break
functionality and indeed the length would have to be longer for this to work
with clang.
How does it work? The code (which you will see in [dds.c](dds.c)) does a `return
system(q-6);` which equates to `return system("cc a.c");` but clang by default,
@@ -79,13 +79,12 @@ that the compilation failed with clang (because it didn't use `make` so no
[LANDER.BAS](LANDER.BAS)) due to a return from `main()` without a return value
and the use of functions not yet declared.
Now as noted since the string `s` is complicated and to not
inadvertently mess something up Cody changed that code to `system("make a");`
and then added the right rule to the Makefile. Thus for the alternative version
you now need both `make` and `cc`. If you use gcc you may change the system()
call to what was given in this description to get a more authentic feel. If you
do this you must then run `./a.out` not `./a`! Thus to use the alternative
version:
Now as noted since the string `s` is complicated and to not inadvertently mess
something up it was changed so that the code uses `system("make a");` and then an
appropriate rule was added to the Makefile. Thus for the alternative version you
now need both `make` and `cc`. If you use gcc you can change the `system()` call
to what was given in this description to get a more authentic feel. If you do
this you must then run `./a.out` not `./a`! Thus to use the alternative version:
```sh
make clobber alt
@@ -101,9 +100,6 @@ produced? How does `a.c` relate to [LANDER.BAS](LANDER.BAS)?
This obfuscated program translates BASIC programs into obfuscated
C programs by way of an obfuscated algorithm.
NOTE: Due to minor problems with some strict ANSI C compilers, we
have supplied a more portable ANSI version.
## Author's remarks: