Android application to manipulate sd image from the site

The only thing I noticed the time I took RISC OS files to linux ADFS and back was a minor change in permissions, that didn’t have any affect in use anyway. I think it is fairly complete. Theo knows a bit about this.

I compiled your last link to squach.c and got a lot of errors. Sometimes errors are cryptic and only happen because a prototype isn’t available.

squach.c:1683: error: 's64' undeclared (first use in this function)
squach.c: In function 'main':
squach.c:2341: error: expected expression before 'else'

We probably haven’t got s64 at a guess.
Edit:

typedef int64_t s64;
typedef u_int64_t u64;

Clears that error (rightly or wrongly -haven’t investigated further)
__int64_t is OK too but not __u_int64_t
now this error

*gcc -c squach.c -static
squach.c: In function 'main':
squach.c:2340: error: expected expression before 'else'

aaaa

aaaa

I read a bit of a book on secure C and C++ today. BSD strlcpy and strlncpy are pretty much equivalent of the _s ones. Unfortunately they are not part of !GCC. Probably macs have them. It sounds like the strncpy is buggy at placing the terminating 0 even in its designed capacity.
strdup would use malloc which you want to avoid, so it is back to memcpy with support I suppose.
If 256 byte buffers are all that is used you could AND the lengths(+ 1..4) with 256 as a safety precaution without going to an if statement.
As far as I got squach did compress a large tar this time. I probably had all the times messed up. There was a lot of reports of Newer (-:
I’ll have another crack at it tomorrow.

aaaaa

aaaa

aaaa

aaaa

Not sure what the header biz is about :)

I guess it means you have a lot of variables or functions that are not declared as static so the compiler expects them to be declared as extern in the corresponding source.h files.

aaaa

aaaa

how am i supposed to do this ?

I think the Norcroft amu tool etc will work with both RISC OS type file structures and unix/windows type. If you name your source files eg main/c and main/h, rather than having them named main in c and h directories, then if you transfer over the network, the file main/c will become main.c and so will be ok on other file systems.

aaaa

cc -c -depend !Depend -throwback -ff -fah -o o.squach c.squach
Norcroft RISC OS ARM C vsn 5.70 [12 Oct 2012]
“c.squach”, line 8: Warning: Undefined macro ‘USE_ZLIB’ in #if – treated as 0
“c.squach”, line 143: Warning: extern ‘bits’ not declared in header
“c.squach”, line 143: Warning: extern ‘next_shift’ not declared in header
“c.squach”, line 144: Warning: extern ‘code’ not declared in header
“c.squach”, line 144: Warning: extern ‘c’ not declared in header
[…]

The “header biz” is exactly because you asked for it!
Option “-fh” (highlighted in bold above) asks the compiler to whinge about everything not validly defined in a header file.

“c.squach”, line 380: Warning: variable ‘dst_format’ declared but not used

Obvious. You have a number of variables declared but are not used anywhere.

“c.squach”, line 593: Warning: Old-style function ‘lzw_dec_init’

I think this means you have defined a function like lzw_dec_init() (with no parameter or void) so the compiler thinks you are using the ancient K&R way of declaring functions.

“c.squach”, line 705: Warning: use of ‘=’ in condition context

Somewhere you have if (meh = blah) when the C syntax means you probably meant if (meh == blah)… I have noticed that the RISC OS compiler is better at picking this up than some.

“c.squach”, line 819: Warning: lower precision in wider context: ‘<<’

You may be doing something like meh = 1 << 28 and due to some painful and dumb rules, you may want to specify something like meh = 1ul << 28ul so the compiler is damn certain everything is an unsigned long.

“c.squach”, line 1034: Warning: odd unsigned comparison with 0: ‘<=’

You are trying to compare an unsigned value as “less than or equal to zero” which is technically impossible. Unsigned doesn’t swing negative.

“c.squach”, line 1095: Error: ‘=’: implicit cast of pointer to ‘int’

Don’t know what your code is, but generally to can’t just do meh = &whatever, you need to cast it.

“c.squach”, line 1104: Serious error: Illegal types for operands:

This possibly came from the previous error? If not, then you cannot use the array.subscript syntax for an indirected array pointer. Try the alternative array->subscript syntax.

“c.squach”, line 1569: Error: <argument 1 to ’strtoul’>: implicit cast of pointer to non-equal pointer

As above, you can’t pass any old thing in and “hope it works” or the compiler will boot it out. You may need to cast. Also check you aren’t assigning a *string to a string[] or that sort of thing.

“c.squach”, line 1733: Warning: implicit cast (to ‘int’) overflow

If you are pushing a large number (bit 31 set), you might have wanted unsigned int?

“c.squach”, line 2097: Warning: Format requires 10 parameters, but 12 given

Clue in the message. ;-)

“c.squach”, line 2162: Error: <argument 1 to ’fopen’>: implicit cast of pointer to non-equal pointer

fopen argument one, that should be the filename, no? You may need casting if you want to use a malloc’d block of data as a string.

c.squach: 128 warnings, 24 errors, 8 serious errors

OMFG.

I take pride in writing code (that may be utter poop) but compiles and links with zero messages.
Here is what happens when I build my MIDI module from scratch:

Stamp @.cmhg.ModHeader
objasm -APCS 3/32bit -desktop -Depend !Depend -ThrowBack -Stamp -IC:, -from s.dadebgptch -to o.dadebgptch
ARM AOF Macro Assembler 4.01 (Acorn Computers Ltd) [04 Jul 2011]
cc -c -depend !Depend -zM -I,C: -apcs 3/32bit -Wcd -throwback -fahn c.usb -o o.usb
Norcroft RISC OS ARM C vsn 5.70 [12 Oct 2012]
cc -c -depend !Depend -zM -I,C: -apcs 3/32bit -Wcd -throwback -fahn c.midi -o o.midi
Norcroft RISC OS ARM C vsn 5.70 [12 Oct 2012]
cc -c -depend !Depend -zM -I,C: -apcs 3/32bit -Wcd -throwback -fahn c.module -o o.module
Norcroft RISC OS ARM C vsn 5.70 [12 Oct 2012]
cmhg -throwback -32bit @.cmhg.ModHeader -o @.o.ModHeader
***> HEADER REBUILT
link -m -o MIDI @.o.ModHeader o.module o.midi o.usb o.dadebgptch C:o.stubs 
appendmsg

Anything more than that is a Fail as far as I’m concerned. ;-)

aaaa

The Makefile used by dde can also be used with gcc, or it needs a different makefile ?

Probably¹ needs a different one, but this isn’t a bad thing – you can use this to pull in the RISC OS specific stuff and help minimise the amount of #ifdef stuff scattered around the files.

Example – when I did a command line program for DOS/RISC OS, I found that the simplest arrangement was to clone as much of conio.h and dos.h as I needed to display “nice” output and read the keyboard. In the DOS version of the program (built with TurboC++ so a rather custom setup there), it used the standard conio and dos. The MakeFile for the RISC OS side simply added in “roconio” and “rodos” which provided the expected functions. The only other change was a small #ifdef in the main header file that defined these functions in the absence of them under RISC OS. From there, the same code would then build on both platforms.

¹ You might be able to get Make to do something with the alien makefile, but I think most people tend to prefer AMU in preference to Make.

“c.squach”, line 593: Warning: Old-style function ‘lzw_dec_init’

I think this means you have defined a function like lzw_dec_init() (with no parameter or void) so the compiler thinks you are using the ancient K&R way of declaring functions.

There are times I feel I’m starting to understand, and spotting a warning about old so called K&R function declarations is one such. The new style has advantages and this warning is one of them:

“c.squach”, line 2097: Warning: Format requires 10 parameters, but 12 given

Clue in the message. ;-)

Anything more than that is a Fail as far as I’m concerned. ;-)

I tend to work on the ideas that:

1. An error or warning message is always telling you you should something different
2. Even if it didn’t go totally t**s up this time you’re pushing your luck should any small variable change.

Maybe I’ve been in support too long but I regard errors and warnings as things that should not happen.

aaaa

aaaaaa

aaaa

aaaa

I didn’t know the norcroft compiler had an option to work with .c .h files.
But anyway, the !Rename application can do the conversion process.
I made a new pattern for going sfix2unix. The unix2sfix is already there.
Edit: The pattern I’m using for sfix2unix is
Prototype ^. Rename To <updir>&1/&f
Set updir ^. (From the CLI or in the Runfile)
There may be another way, I didn’t find it.
The Makefiles are different, and though the norcroft has a filetype for a Makefile, The system seems to
filetype the gcc text Makefile sometimes. Probably easiest to use a different name for the gcc make makefile and call it with make -f linuxmkfile or something.

I didn’t know the norcroft compiler had an option to work with .c .h files.

Hmmm – maybe I have extrapolated too far. It will accept either form of source file path in , for example, includes, i.e. #include “main.h” or #include “h.main” both work. In some of my source files there are both types in the same source file. Bad practice, but resulted from copy/paste actions between source files.

OK so it /must/ have c directory and h directory. I can recall being confused a little by seeing ambiguity in the makefile.
Here is a simple gcc makefile for squach. You could add in more CFLAGS for more stringent error checking.

SHELL = /bin/sh
TARGET = squach
CC = gcc
CFLAGS = -static
LIBS = -lz
OBJS = adfs_time.o blocks.o strings.o squach.o tar.o lzw.o
$(TARGET):	$(OBJS)
	$(CC) $(CFLAGS) $(OBJS) $(LIBS) -o $(TARGET)
clean:
	rm -f $(OBJS) $(TARGET)

aaaaa