TIGCC 0.96 Beta 7

RSS
Email

0 Kudo

Author	Comment
Kevin Kofler	TIGCC 0.96 Beta 7	Lead	[-]
Posts: 7059 (2006-07-30 07:45:01) Admin Reply Quote More My Recent Posts Message Me Connection Blocking Invite Report Post	I have just uploaded TIGCC 0.96 Beta 7 for Windows and for Linux/Unix/OS X. Thanks go to Frédéric Bour (Defre) for recompiling ide.exe and tigcc.exe and Joey Adams (MrJoey) for rebuilding tigcc.chm. TIGCC 0.96 Beta 7 Release Notes: Changes (since Beta 6): * TIGCCLIB: gray.s size optimizations by Jesse Frey * TIGCCLIB: changed jsr to jbsr in trampoffset.s * TIGCCLIB: merged a few more of Lionel's documentation updates (edited by Jonathan Holbrook) * A68k: fixed compilation on recent versions of BSD and OS X GCC: updated to 4.1.2-tigcc-pre8 (based on the 4.1-20060728 prerelease snapshot). * ld-tigcc: The standalone version supports a new --outputbin-main-only switch to be used by KTIGCC. * TIGCC IDE: MOV is not an A68k keyword. * TIGCC IDE: fixed duplicate/missing keyboard shortcuts in the Project Options, Program Options and Preferences dialogs * TIGCC IDE: fixed MLink format relocs radiobuttons not being disabled for kernel programs in the Program Options dialog * TIGCC IDE: also search Include\S for include files with errors * TIGCC IDE: don't show Debug menu for FlashOSes * ExePack support is now Free Software (LGPL + exception to allow distribution of unmodified binaries without source code for pstarter and ttbin2oth, plain LGPL for ttpack) * TIGCC for Windows: MSVCRT.DLL is now required. This DLL comes with all Windows versions since Windows 95 OSR 2. For the original Windows 95, it is usually installed by Internet Explorer or Microsoft Office. * TIGCC/nix: split out the TIGCC Tools Suite into a separate package TIGCC/nix: fixed build scripts to use gmake instead of make on BSD TIGCC/nix: fixed install script not to attempt symlinking os.h to OS.h on case-insensitive file systems TIGCC/nix: fixed ld-tigcc compilation on XCode 2.3, added note that XCode 2.2 and 2.2.1 are not supported TIGCC/nix: Linux/x86 binaries now built on Fedora Core 5 (glibc 2.4) TIGCC/nix tigcc: fixed external data variable support for compressed programs TIGCC/nix tigcc: fixed multiple uses of -Wa or -WA (fixes pedrom-ld-tigcc build) tprbuilder: added -D switch (forwards defines to tigcc) (Lionel Debroux) Important notes: * The "Generate debugging information" option in the IDE and the "-g" switch in the command-line compiler work differently starting from TIGCC 0.96 Beta 4. They no longer produce human-readable annotated .s files for debugging programs by hand using an assembly-level debugger. Instead, they produce .dbg files containing debugging information in the binary DWARF 2 format, which can be used with GDB, which will be integrated in TiEmu 3. A snapshot of TiEmu 3 is available at: http://sourceforge.net/project/showfiles.php?group_id=23169 * The nix release no longer contains the TIGCC Tools Suite. It can be obtained separately as a tarball (source and binaries) or RPM from: http://members.chello.at/gerhard.kofler/kevin/tts140/ (The plan is to upload these to ticalc.org, but I need to check with Lionel because this is a preliminary version, but he wanted me to package this instead of the very old 1.31.) Still to do (planned for TIGCC 0.96): finish the Linux/Unix IDE (ktigcc) - Summer of Code project running, usable alpha releases available * merge Lionel's (remaining) documentation and library updates into TIGCCLIB * merge my 3-plane grayscales into TIGCCLIB (assigned to me, of course) * support for 89u/9xu/v2u files (not just tib) in the linker * implement GNU as .[su]leb128 and .cfi* directives (in a way compatible with linker optimization), in order to facilitate writing debugging information for ASM files. * add DWARF 2 function prototype and unwinding information to tigcc.a Still to do (not scheduled yet, may be ready for 0.96 or not): * work on having GNU as accept A68k code, so we can drop A68k and maintain only one assembler. (Some things are already handled by -mri-compat, some more things are easy to handle, others are harder.) Still to do (not scheduled yet, will most likely not be ready in a 0.96 timeframe): * support for multi-section programs in the linker as suggested by JimRandomH * LZMA compression support (not high priority anymore now that ExePack is Free Software) [EDIT: Release note formatting fixes.]
Lionel Debroux	Re: TIGCC 0.96 Beta 7	#1	[-]
Posts: 3065 (2006-07-30 14:11:41) Former Admin Reply Quote More My Recent Posts Message Me Connection Blocking Invite Report Post	Of course, thanks a lot for processing and adding some of my additions to TIGCCLIB ! I guess some of the ~30 "easy" files I told you about a while ago are among the processed ones (I've downloaded Beta 7, but it is not installed yet on the other machine) ? I can remove the modified tprbuilder from TICT-Explorer, since it is part of the official distribution. I have worked a bit on suppressing the warnings / errors from the TIGCC Tools Suite pctools. So far, I suppressed most of the easy ones (free unsigned / signed pointers mix, comparisons between signed and unsigned), and some of the 64-bit incompatibilities. And I don't forget that I had started small modifications to the TIGCC linker, so that the timestamps in the OS it generates are correct.
ZPedro	Where?	#2	[-]
Posts: 55 (2006-07-30 16:27:48) Reply Quote More My Recent Posts Message Me Connection Blocking Invite Report Post	All right Kevin, where do I get GCC 4.1-20060728? Don't tell me a GNU mirror: even mother ftp://ftp.gnu.org/ does not have it (the lastest there is seems to be gcc 4.1.1 from 25/05). Unless I'm missing something... (no, I won't get that version from anonymous cvs, or svn for that matter) Anyway, I don't think you should be using such prerelease snapshots: mirrors tend not to pick them up, or remove them after some time. Heck, two betas ago when you first put a bunch of fixes for OSX after I reported them, I could not find the accurate prerelease one you told to use but only a version that came shortly afterwards, fortunately not patch got rejected except the version one. This time there isn't even a version after that snapshot that I could more or less reliably use.
Tyler	Re: Where?	#3	[-]
Posts: 47 (2006-07-30 16:37:19) Reply Quote More My Recent Posts Message Me Connection Blocking Invite Report Post	Great! I will download this right away
Kevin Kofler	Re: Where?	#4	[-]
Posts: 7059 (2006-07-30 20:11:16) Admin Reply Quote More My Recent Posts Message Me Connection Blocking Invite Report Post	gd.tuwien.ac.at/gnu/gcc/snapshots/4.1-20060728/gcc-core-4.1-20060728.tar.bz2 It's also on the gcc.gnu.org FTP, you need to look in the snapshots directory. And when doing a TIGCC release between 2 GCC releases, I only have 4 options: * wait for the next GCC release (in this case 4.1.2) to publish TIGCC - not that great an idea, it holds me hostage to GCC's release schedule * use the previous GCC release (e.g. 4.1.1) - not that great either, you don't get the current bugfixes on the 4.1 branch * use snapshots - that's what I'm currently doing * use SVN checkouts - that's even worse (but you always have the option to get a SVN checkout from the date the snapshot was taken at - for the time, use something around 2230-2245 UTC) What I should do instead is make sure my stripped-down GCC tarball contains the host support files for all supported hosts (that would be at least all Linux, BSD and OS X platforms, probably more nix hosts). The problem is that GCC mixes up host and target support files all in config/, so it needs careful cleaning to remove only the target support crap I don't need and keep the host support files. What I can try to do in the meantime is to provide a diff from gcc-core-4.1.1 to gcc-core-4.1-20060728.
ZPedro	-	#5	[-]
Posts: 55 (2006-07-30 21:06:25) Reply Quote More My Recent Posts Message Me Connection Blocking Invite Report Post	I see your point. But you're doing a release, which means it may be used 6 months from now on and generally be more easy to get. Most mirrors don't have this snapshot directory. If you're worried about the bugs fixed since the last gcc release, then why live at the bleeding edge? This edge hurts you know, you could settle for the stable branch (4.0.x). If you really want to benefit from the development of the development branch then you should indeed provide a way to get this snapshot more easily. Anyway, I'll be off the 'Net for the whole of August. See ya then.
Kevin Kofler	Re: -	#6	[-]
Posts: 7059 (2006-07-30 21:33:24) Admin Reply Quote More My Recent Posts Message Me Connection Blocking Invite Report Post	The bleeding edge would be "4.2.0 20060729 (experimental)", not "4.1.2 20060728 (prerelease)". Linux distributions such as Fedora and Debian also use SVN snapshots from the branch for their packages, not releases, and Apple does so too. Really, the problem is not that I'm using a snapshot, it's that getting a copy complete with all host support files may be hard a few weeks/months from now, so this is the problem to be solved.
pinh3ad	Re: GetDataType	#7	[-]
Posts: 59 (2006-08-02 16:53:23) Reply Quote More My Recent Posts Message Me Connection Blocking Invite Report Post	Since I use TIGCC v0.96 beta 7, there are 3 warnings in one of my projects: - 'Implicit declaration of function "_row_call_addr".' - 'Cast to pointer from integer of different size.' - 'Cast to pointer from integer of different size.' All these warnings are caused by the following function: // returns the type of a variable as a string (in buffer)void VarTypeToStr(SYM_ENTRY var, char buffer){ char *tag; tag = HToESI(var->handle); if (GetDataType(tag) != SDT_OTH) { strncpy(buffer, SmapTypeStrings(GetDataType(HToESI(var->handle))),4); } else { tag -= 5; // 3 letters? if (!isalpha(tag[0])) tag++; strncpy(buffer, tag,4); } buffer[4] = 0;} I have no idea why there are any warnings, because everything seems to work fine! Is this a bug of TIGCC or what occured? [EDIT by Kevin Kofler: Disabled emoticons.]
Kevin Kofler	Re: GetDataType	#8	[-]
Posts: 7059 (2006-08-02 21:03:51) Admin Reply Quote More My Recent Posts Message Me Connection Blocking Invite Report Post	This is due to a typo in vat.h. In the definition of GetDataType (towards the end, line 137), change _row_call_addr to _rom_call_addr. Blame Jonathan Holbrook for this one. I just double-checked Lionel's submission and it's correct there. Discussing this with Jonathan, we have no idea how that error crept in.
Kevin Kofler	Re: GetDataType	#9	[-]
Posts: 7059 (2006-08-02 21:13:50) Admin Reply Quote More My Recent Posts Message Me Connection Blocking Invite Report Post	Fixed in CVS. tigcc-linux.cvs.sourceforge.net/tigcc-linux/tigcc/include/C/vat.h?r1=1.4&r2=1.5 tigcc-linux.cvs.sourceforge.net/tigcc-linux/tigcc/doc/System/Include/vat.h/GetDataType.hsf?r1=1.2&r2=1.3
Kevin Kofler	Re: GetDataType	#10	[-]
Posts: 7059 (2006-08-02 21:19:45) Admin Reply Quote More My Recent Posts Message Me Connection Blocking Invite Report Post	As for why these warnings don't affect you, the code they affect is only ever used if MIN_AMS is <200. Lionel Debroux added some code to make that particular function work with a lower MIN_AMS setting. Jonathan merged it and that error went in in some way I don't understand.
mhubi	thanks	#11	[-]
Posts: 126 (2006-08-03 11:45:15) Reply Quote More My Recent Posts Message Me Connection Blocking Invite Report Post	i want just to say: great thing, that there's a new version and people know, this project is still alive. thank you very much for your work.
Lionel Debroux	Re: thanks	#12	[-]
Posts: 3065 (2006-08-03 13:15:52) Former Admin Reply Quote More My Recent Posts Message Me Connection Blocking Invite Report Post	What about writing the function the following way (untested): void VarTypeToStr(SYM_ENTRY var, char buffer){ char *tag; char type; // Maybe a short, I don't remember... tag = HToESI(var->handle); if ((type = GetDataType(tag)) != SDT_OTH) { strncpy(buffer, SmapTypeStrings(type), 4); } else { tag -= 5; // 3 letters? if (!isalpha(tag[0])) tag++; strncpy(buffer, tag,4); } buffer[4] = 0;} Note that in order to make safe code, you should also fix all possible dereferences of NULL: HToESI returning NULL and subsequently dereferenced by GetDataType; SmapTypeStrings returning NULL (I guess it does if the type is invalid, but it shouldn't happen here). [EDIT by Kevin Kofler: Disabled emoticons.]
pinh3ad	Hm	#13	[-]
Posts: 59 (2006-08-03 15:27:25) Reply Quote More My Recent Posts Message Me Connection Blocking Invite Report Post	Hm, yes, that would be better. And you are right, I should really check if HToESI returned NULL. Thanks for the advice!

<< Previous Topic Next Topic >>

Yuku free message boards

TIGCC 0.96 Beta 7

TIGCC 0.96 Beta 7

Quick Reply