TungsterDev compiling

I’ve downloaded TungstenDev sources and had these problems.

UntarBz2 off the web site doesn’t work it clips the end off long filepaths

Can you replace the ‘Tar’ program inside untarbz2 with tar.zip

When compiling the sources it fails with

"TCPIPLibs:sys.h.errno", line 95: Error: differing redefinition of #define macro EFBIG

This may be a AcornC/C++ problem as the copy of sys.h.errno in there is identical. My copy of AcornC/C++ is recent.

Anyway I put

#undef EFBIG

before the definition of EFBIG in that file and everything compiled ok.

ps specialstrings work ok with the new rom.

Anyway I put #undef EFBIG before the definition of EFBIG in that file and everything compiled ok.

Unfortunately that’s a rather simplisitic change in this instance. The TCP/IP libs define EFBIG as 27 (since they’re derived from BSD) but the one in the CLib defines it as 6.

So a program like

int main(void)
{
  int result = SomeTCPFunction();

  if (result == EFBIG) DoSomething();
  return 0;
}

wouldn’t work as the library function would return 27 but the test would be against 6.

It’s further complicated that the compiler contains built in copies of the standard ANSI headers (for speed) so even ammending the disc copy wouldn’t help. The compiler needs updating too, or the magic -j switch used to force the headers to be read off disc.

Since Linux, NetBSD, Borland C, FreeBSD all use 27, for my money I’d change the CLib definition as that only got added in the last 18 months, but then the compiler is out of sync.
The nightly builds are done with cc 5.69 so don’t trip up.

Was there just one or two things tripping up? I imagine they must be sources that include both <errno.h> and TCPIPLibs:sys.h.errno are they? They could probably be reviewed and fixed up.

I’m using 5.70. Rool want to buy themselves a copy :-)

I’ve recompiled and these are the problem files.

The EFBIG error is in

inetlib
c.inet_pton
c.inet_ntop
c.inet_npton
c.inet_nntop

I think there should only be one value. You don’t know which one a library uses so can only guess.
I figured that TCPLibs would use the value in sys/errno.h so bodged that but it obviously doesn’t work as the value depends on which headers are used. Presumably changing sys/errno.h to be 6 would make everything the same without changing the compiler – which is fine except 27 may be the prefered value. I’ll let someone else sort it out :-). I’m still on the internet so so far so good.

I’ve also just found

colourtrans 
MkTables.maketables @
Constructing 8bpp desktop tables
  - written palette.
  - written table.
Constructing 8bpp grey tables
  - written palette.
  - written table.
Constructing 4bpp desktop tables
  - written palette.
  - written table.
Constructing 4bpp grey tables
  - written palette.
  - written table.
squash Tables.4greys
Squash: Invalid Squash header
AMU: *** exit (255) ***

The EFBIG error is in inetlib

Looking at the head revisions of pton and ntop in FreeBSD they’ve removed sys/errno.h and just stuck with <errno.h> so that points to changing EFBIG to be 27 throughout. At least if it’s only inetlib that clashes the corresponding makefile can be changed (for force the compiler to use the disc copy of the headers) until ROOL get round to updating the compiler.

I’ve also just found ‘Invalid Squash header’

You didn’t do a clean build before exporting resources. That is fixable, but we’re all build-from-clean saints aren’t we?

You didn’t do a clean build before exporting resources

Who me! In my defence m’lud I was only doing a quick recompile to find errors – honest. Promise not to do it again.