Does anybody know where I can have the sources ?
It is a GPL licence and I can't find the sources !

Sorry for this comment, i have hust found them.

Ok, It's me again.
Where i can found the source of gcc ?
(I don't find them in tigccsrc.zip)

A patch to the original GCC is in the gcc subfolder in the file tigccsrc.zip; you should have found that. It contains the differences between my files and the original GCC sources. These can be found at www.cygwin.com (for Windows, otherwise look at www.gnu.org/software/gcc, although they should be exactly the same). You should be able to use 'patch' on this patch if you change all the directories in the patch, although of course I haven't tried it. If you know how to use 'diff' properly to generate a patch in the format needed by 'patch', please tell me, because I haven't used any Unix or GNU tools before.

If you absolutely need the whole sources and Cygwin installation (which is about 300 MB, I think), I could send you a CD.

If you plan on compiling for Windows, click the "Install Cygwin Now" picture on www.cygwin.com. The sources can be found if you go to the mirror sites section, because they are on the FTP.

Hi all,

Is there a working linux port of the TiGcc ?
Grabbing the source of GCC and patching with the included patch doesn't work for me.

> Is there a working linux port of the TiGcc ?

mmu_man made one, but he never released it. The presently uploaded one doesn't change the return value convention.

> Grabbing the source of GCC and patching with the included patch doesn't work for me.

Why not? Is there something wrong with the patch? I really don't know anything about making patches, so please tell me exactly what I did wrong or what I have to do better. In particular, all the patches I have seen so far always included at least two more lines above and below each change, but I don't know how to instruct 'diff' to do that, since I don't have any docs on 'diff'.

Oh, and remember to configure with '--target="m68k-coff"' first, from another directory.

I have Linux (actually, a dual boot of Windows and Linux). I was seriously considering investing in Borland's Kylix (which is basically C++ Builder and Delphi for the Linux platform). If I did that, I would be able to compile the IDE under Linux, and that would solve that. As for GCC, I am sure that would not be to hard to do either. Anyhow...

That's about my $0.02 worth for today.

> I was seriously considering investing in Borland's Kylix (which is basically C++ Builder and Delphi for the Linux platform).

AND Windows! :-) I would do the same, if only I had enough disk space and money and time and ...

> If I did that, I would be able to compile the IDE under Linux...

Yes, but be aware that the IDE uses a lot of Windows-specific stuff directly (sorry). Converting it should be possible, though.

> As for GCC, I am sure that would not be to hard to do either.

It was hard for me to even apply the TIGCC for Linux patch to the original GCC when I was trying to dive right into it. But if you already know Linux, you are probably familiar with patches, makefiles, and the like.

Actually, I am fairly new to Linux. Sebastian, I haven't looked at the IDE's source. There is nothing wrong with making API calls. He||, I do it all the time. It makes it easy to do something hard (most of the time). Anyhow... :-)

Wow :)
3 weeks break and the world still turns round ;)
Sorry but I was very busy this month (I just finished my exams, I'm on holidays now for nearly 2 weeks).

I'll reply to the whole message tree in order :

endian: Yes, and I do use it to compile the Prosit kernel, so it works, belive me ;)
No, the tigcclinux package on ticalc doesn't work (see reason below).

Sebastian: Yes I didn't had time to make a package... but downloading the tigcclinux package from here, the gcc and binutils packages, and following the instructions from Henri Moilanen using my patch instead of his patch actualy works. (and compiling the last pseudo-linker from the sources of tigcc-win).
Henri Moilanen's patch doesn't work because it doesn't change the return convention (pointers in a0), though it should be the main thing it ought to do...
mine fixes this.
Doc on diff: use the man page :)
see http://linux.com.hk/man /showman.cgi?url=http://linux.com.hk/man /man1/diff.1.inc (remove the space in 'man /' twice)
(btw, I'd like to have the diffs from tigccwin instead of the whole 300Mb sources)

cajunguy: I'm not sure where the point of building tigcc IDE under Linux is...
1) You have great tools like (X)Emacs http://www.xemacs.org/ or even Kdevelop (didn't try it, I still stuck on emacs), but of course, the more choice you have, the best it is ;)
2) You still can try running it with Wine http://www.winehq.org (.com ?) or even recompile it using the wine libs.
I tried the IDE a long time ago and really felt bad with the bare slow syntax hilight, but I heard it is fixed now :))) (btw I still stuck on emacs, since I now exclusively work (I said _work_, can you do that under wincrash ?) under Linux)

Sebastian:
Y-ou still have the option of using the wine libs which are a rewrite of windows API under Linux :)
-I thought it was very difficult, and in fact all you have to do is patch -p0 mmu_patch :) (of course in the good directory)

> Yes, and I do use it to compile the Prosit kernel, so it works, belive me

The present Prosit kernel unfortunately crashes my calc. :-(

> and compiling the last pseudo-linker from the sources of tigcc-win

Or use mlink, from alh.dhs.org/ti89.

> Doc on diff: use the man page

Thanks. So I guess I have to use the -c or -C option. I'll try that.

> You still have the option of using the wine libs which are a rewrite of windows API under Linux

I'm sorry it's not quite that easy. Especially calling the programs heavily depends on things that probably can't be emulated. I don't have Linux, mainly because of a lack of disk space, so I can't try this.

> I thought it was very difficult, and in fact all you have to do is patch -p0 mmu_patch

I didn't have any problems with your patch; it's just hard to find the sources recompile GCC if you have never even heard of GNU before. :-)

BTW: You might even be able to use the present patch from the sources, but only on the most recent version of GCC.

> prosit kernel
You don't have the last version ;) (btw it works, but on Vti, since I don't have a 92+/89, only a 92, and I still wait for [tie] to come in a new version to see Prosit in real :))) ) (also, I ment it works as the windows compiled one :p )
Check the url below ;)

> mlink
didn't know about mlink, I'll check it.
Anyway, I'll try to implement some kind of 'archived libs' i.e. a library which is executed _in the flash rom_, with no need of relocation, I think it is feasable with the help of a normal (wrapper) library which searches for the alib, then calls the function. (searches once and get a pointer to the function table)
I don't expect I'll be able to make such libs in C as the coff doesn't handle PIC code well, but mabe one day (the uCLinux port uses a pic-coff-gcc for the user-space progs)

> diff
As I remember you are on the way :)

> wine
Well I managed to run even WinAmp with Wine (IE still sucks because my version of Wine has a broken networking support, maybe in the last version)

> gcc
Yes I follow you :)
When I first compiled gcc (for tigcclinux) I was frightened, and I still don't understand everything... I'm currently using makefiles for Prosit, but nothing beyond that (./configure... I wouldn't know how to make this, I heard it wasn't so difficult though)

> patch
Yes, I saw the message above :)
I didn't know you put the patchs in the package :)

concerning the nostub and the asm libs question I found somewhere, I wrote a small header file to use DoorsOS' filelib in C, and it works well, I'll try to release all that stuff soon ;)
(along with a full tigcc for Linux)

Some project in my mind :
- Prosit:
- VFS,
- networking layer (TCP/IP, ...)
- header files for the asm libs
- tigcclinux
- modification to 'link' to be able to create Fargo progs from tigcc (so I could test Prosit without [tie], and it could help porting other progs too-thinking of tichess even if it is memory hungry)
- TCP/IP and file libs (POSIX like), as well as other stdlibs
... don't know if I'll be able to do it all ;)

>I'll try to implement some kind of 'archived libs' i.e. a library which is executed _in the flash rom_, ...

Sorry, I think you misunderstood this when reading it: As far as I know, when talking about "archived libs", Zeljko Juric and Sebastian Reichelt meant something completely different. They were talking about STATIC libraries containing multiple object files, from which the linker would extract only the useful ones, in order to no longer have to use the hex array functions in TIGCCLIB.

Executing anything from Flash ROM is IMPOSSIBLE except if you manage to break the protection by using some very dirty hacks. I personally don't know anyone who succeeded in this.

It was in fact Xavier Vassor who wants to do this, but he can't even fix the bugs in his linker. mlink already supports this, and it is better anyway, but Johan doesn't want it included in TIGCC, nor does he want to release his source until version 3.0. I really wish we could include that linker.

Sorry, there is a mistake here, I do know what I'm telling, I'm not talking about _real_ libraries (those you describe, as used by every __normal_ compiler, except tigcc-BTW, I'll try it too, as under Linux I have m68k-coff-ar, so I can).
Sebastian> maybe instead of wanting Xav's linker to support it you should consider using a _real_ linker (the one which comes with the compiler in the binutils, m68k-coff-ld, which handles multiple objects fingers in the noze) Of course it can rise the problem of using the libraries with asm programs, as they aren't in the coff object format (BTW I still think that chousing the amiga object format for gcc would have led to less problems...)
What was I telling ? Oh, yes, I was talking of flash libs that could run in the flash (without having to be relocated). Of course there is this stupid protection I forgot for a while... but Hey, nothing is impossible ;)
Sebastian: talking about mlink you sait there are no sources ? How will I use it under Linux ???? Closed-source progs REALLY SUCKS :-(

mlink: On alh.dhs.org/ti89, there is both a DOS and a Linux version available.

m68k-coff-ld: Hey, thanks to the MAN pages, I now know how to do this: with the -r switch. This is the way it will be done in the next release. Thanks!

BTW, Xavier's linker _is_ a real linker; it's just buggy.

amiga format: How would this be done? Is this the A-Out format? Please tell me the exact target name (like m68k-coff).

> mlink
ok, but what if I had an Alpha ? I wouldn't be able to recompile it without the sources... x86 segregation...

> ld
:)

> link
Well, it used to be only a convertion tool, but it's becomming a linker (the last version I tested wasn't linking more than 3 objects correctly, but maybe I should try the last one :) )

> objects
don't remember the exact name, (in fact there is more than one), one of m68k-amigaos, m68k-cbm, m68k-amiga
It seems it is no longer supported in the last gcc versions (since the egcs story-the target name exists, but the ./configure says it's not supported-, but maybe it will be a supported target again)

Hey mmu_man, can you just send me your Patch ?

Ok, it's me again ... and i'm need of help.
I can't still get the Gcc compiled. The patch applied with some newline-problems, if you can call it so. I'll explain the problem if you want to but they are easy to solve.
Now I can compile the binutils without any problems with the m68k-coff-target. When I try to compile the gcc with the applied Patch, it stops at some line saying something of xlimits.h, looking at the Makefile it seems to me that this is ok, because there's a comment saying "We only need the C Compiler so stop here". Then I do a make install and he tries to compile some stuff stopping with the line 'g77 requires ability to access exact FP representation of target machine'. I don't know what to do,if you could help me ?

Perhaps you could get me some instructions on how to compile the whole tigcc ?????

Thanks in advance...

---
sorry for my bad english, it's not my natural language.

FINALLY! Floating point support! I don't have to cope with TI's "ti_float"form of vairable and all that flt crap anymore.

I have added it on my CDrom

Isnt that illegal?

It _is_ ! It is Free Software :p

Actually, it's not illegal to do this.

From the GNU General Public License:

When we speak of free software, we are referring to freedom, not price. Our General Public Licenses are designed to make sure that you have the freedom to distribute copies of free software (and charge for this service if you wish)

-Sean