Conference hydra::amiga_v1

This is unrelated to .1.

Lattice has a version 5.02 patch comming out. It will be sent to 
all registered owners. (anyone ever seen 5.01??)

I don't know the details, but know of at least one fix - LSE.

There is a bug in LSE that when you resize the window, then compile
a C file from within the editor, it gets all weird. Sometimes hanging,
usually just prints a row of garbage across the "status line".

Actually it's not "garbage" at all. It consistantly prints "jGjGjG" across
the whole status line. Well those just happen to be my initials! And here I
thought it was a tribute to my first successfull compile from within the editor!

John

Yes, I have Lattice C 5.01. I got the patches from the Lattice BBS just
before Christmas. I fixed TONS of bugs. It enabled me to finally get
Sedt working. 5.0 has so many bugs the Sedt fell all over itself. I
found quite a few using Codeprobe and submitted them to Lattice. They
fixed all mine in 5.01.

By the way, anyone use the global optimizer that comes with 5.0? I
ran it against Sedt and although the executable was now a bit smaller,
Sedt crashed immediatly!

Oh well, waiting for 5.02.
Frank

Well, wait no more. 5.02 is on the BBS!  I have downloaded only the 
502READ.ME And the LSE.ARC patch so far. Just tried to get back on
and there is no answer (off line or too many users?). The LSE.ARC
contained 1 file named LSE.PCH which is niether an executable or a
script (EXECUTE) file. Lattice [the rest of the 5.02 patches] must have
some patch program that uses this as a data file. 

Anyway, here is the end of the READ ME file (first part is original 5.0 stuff)

John

(502read.me)

	.
	.
	.

Problems corrected in V5.02:
===========================
 
LC1 and LC2
-----------
1)  Several CXERR's have been fixed.                
2)  Incorrect code generation for logical comparisons has been fixed.   
3)  Failure to promote unsigned for "-w" has been fixed.
4)  Functions loading doubles off the stack from incorrect offsets with 
    "-f8" has been fixed.
5)  Registerized parameters loads up parm with correct offsets.      
6)  Short integer constants too wide for a short with "-w" option are now 
    silently promoted.                       
7)  Formal register double parameters are now loaded correctly with "-f8".
8)  Mixing "-b0" and "-r1" options now works correctly.         
9)  Undefined symbol errors have been fixed.                  
10) Fast floating point problem, "-ff", has been fixed.
11) Mixing "-b0" and "-x" problem has been fixed.        
12) The "-b0" option generating incorrect offset for reverse pc-relative 
    calls has been fixed.
13) The __saveds keyword has been fixed.                          
14) 68020 generates correct code for indexing.  
15) Address of external strings coalescing with "b0" has been fixed.
16) The "-v -y" options now work correctly when used together.
17) The "-x" option now works for initializers.
18) Listing option puts correct version number on page headers.
19) The __builtin_strlen function has been fixed.
20) Chip keyword no longer causes relocs to incorrect hunk.            
21) The volatile keyword now works.                
22) Autoincrement with shorts now clears out data register.  
23) Indexing by an unsigned short no longer loses register contents.
24) Insufficient registers for call through function ptr has been fixed.
25) Header files causing a stack overflow has been fixed.           
26) Constants like 0x3E8 are now loaded correctly under ms.         
27) Multiply by 255 now works.                           
28) The "-s" option now works.                                 
29) Compiler now preserves D2/D3 on calls to CXM55.      

CodeProbe
---------
1)  The catch command no longer causes gurus if an invalid task name is 
    specified.
2)  Detach successfully restarts a task before detaching.
3)  If a task is waiting on a message port when a <CTRL>-C is typed
    from the debugger, the task can now be re-started successfully.
4)  The disassembled instruction bytes are now displayed as words:
    00C01234: xxxx xxxx xxxx      MOVE.L  a0,foo
5)  Invoking the debugger before any other access to ram: no longer causes
    the debugger to hang.
6)  The commands dd and df with the %e and %g format now print exponents 
    correctly.
7)  The dump commands now display correctly when a register is given as
    an argument.
8)  The debugger now removes tasks that exit by calling RemTask(0)
    from its active task list.
9)  If a call to DeviceProc() or OpenDevice() open a device for the very
    first time, an AddTask() is performed to create the Device handler.
    By default, the debugger no longer grabs the new task. In most cases 
    this is the desirable action, however this is probably not what people 
    who debug device drivers will want to do.  Therefore, an option is 
    provided to allow these tasks to be caught. This option is the last
    selection in the Options menu.
10) AutoSwap mode and instruction bytes now default to OFF.

LC
--
1) "H" option fixed.
2) "q" or "o" option lets you specify any extension for precompiled headers.
3) "R" option fixed.
4) Put "" in .lnk files so filenames with spaces will work.

BLINK
-----
1) Handles "" in with files correctly.
2) Handles Overlays correctly again.
3) Handles large libraries and libraries with debugging information 
   correctly.
4) Does not write out 0 length code hunks.

LSE
---
1) Compile from memory gives you more information and does not cause
   bad characters to appear on the bottom of the screen.
2) The autoindent and '}' bug has been fixed.
3) Interlace mode now sizes window correctly.
4) The "-o" option is no longer necessary on compile options.
5) LSE is significantly smaller in code size.
6) You can now RUN LSE and still compile from memory.
 
LMK
---
1) Handles default rules correctly.

ASM
---
1) Minor code generation bugs fixed.
2) Puts out warnings for BRA.L and 68030 instructions on 68000.
3) CSECT allows PC relative or base relative addressing by default.

TB
--
1) Handles Chip and Fast bits correctly.

LIBRARIES and HEADERS
---------------------
1) lcm881.lib works correctly now.
2) Stubs added for registerized parmeters so AMIGA.LIB is totally optional.
3) _exit fixed.
4) Overlay manager fixed.
5) The startup routine cback.o has been fixed.
6) DeleteExtIO and CreateExtIO now are like AMIGA.LIB. The prototype in
   proto/exec.h was also corrected.
7) astcsma is now case insensitive.
8) A bug in DOS.H that happened when short ints were used has been fixed.
   It fixes the problem with getmem() returning memory on long word
   boundaries only, and fixes the problem with getfnl().
9) A prototype for system() was added to stdlib.h
10) CreateTask and DeleteTask now function like AMIGA.LIB
11) LCNB.LIB and LCSNB.LIB now work without linking in the math library.
12) The macro offsetof was added to stddef.h
13) BSTR * was changed to BSTR in libraries/dosextens.h
14) The %e in printf and similar fuctions has been fixed.
15) grep.lib included.

    Is patch  5.01 needed before 5.02 patches?  Or can I just down load
    the .02 patches?
    

Not sure. But seeing how Lattice is sending 5.02 to registered users and
has never sent 5.01, I imagine not. 5.02 *probably* has the 5.01 fixes as well.

I couldn't get to the BBS all weekend so all I have is the READ.ME and
a (useless?) LSE patch.



John

P.S.  Mr Moderator, could you please move all these replies on Lattice to their
own note with a more apropriate title??

Re .3:

>script (EXECUTE) file. Lattice [the rest of the 5.02 patches] must have
>some patch program that uses this as a data file. 

John,

On disk 3 there is a program called PATCH which you use to install
the patches.  I haven't used it but I read about it in one of the readme
type files on one of the disks (can't remember disk or filename).  The
command format is

PATCH  oldfile [-ooutput -ppatchfile]

where oldfile is the old file to be patched.
      output is the output file, default is oldfile.new.
      patchfile is the patchfile supplied by Lattice.

Rich

P.S. Thanks for entering the list of corrected problems.

Yeh, I got blasted big time by Carlotta (SYSOP) on the Lattice BBS last night!

Apparently She had posted a note on it (one of 5000+). I had just figured out
how to get around the BBS and was rather pleased to have left a "public"
message.  Silly Me!


Thanks Rich!

John