home *** CD-ROM | disk | FTP | other *** search
- ----------------------------- PatchLevel 8 ----------------------
-
- atarist.c: andreas
- - allocate kludge env after counting amount required, free after
- child exits
- - relocate misc functions.
-
- st-traps.s: andreas
- - change all those ugly sequences of movel's to moveml
- - when testing for longframes, use "tstw 0x59:w"
- - change a few jmp's to bra's
-
- thanks to andreas and michal for their diffs/testing/efforts
-
- atarist.c:: andreas
- do regexps
-
- atarist.c:: ++jrb
- fix up debugee exit problem: save the context before the pexec,
- and restore it when child returns before doing anything else.
- now it woks great. before the stack was completely hosed at
- this point.
-
- use _initial_stack instead of _stksize to force mallocs from
- private heap. note this is the size for both stack and heap
- to it better be big enough for mallocs, stack and alloca's on
- stack (gdb does plenty of them). this should also help us
- when (and if) we integrate readline, gdb will now never cause
- a Malloc() gemdos call to occur at in-oppurtune times, nor will
- it get hosed when the child exits.
-
- version.c:: ++jrb
- put Patchlevel in version str. easier to track reports that way.
-
- defs.h:: ++jrb
- adjust a few things.
-
- m-atari.h:: ++jrb
- fix up FRAME_CHAIN, plus a few other adjustments
-
- st-traps.h:: ++jrb
- make all the context stuff volatile beacuse they really are (ie: we
- change them from under the compiler).
-
- ------------------------------- PatchLevel 11 -----------------------------
- (never mind all the Patchlevels in between!)
-
- fake/read.c, fake/lib.h:: ++jrb
- sync with changes in library.
-
- ------------------------------- Patchlevel 12 -----------------------------
-
- atarist.c:: andreas
- the blockvector gets relocated twice if there are subfile entries
- in the symbol table (e.g., inline functions in a header file).
- fortunatly the subfile entries directly follows the main file in
- the symbol table chain, so it is easy to figure out if a
- blockvector is already relocated (just see if it is the same as
- the last one)
-
- ------------------------------- Patchlevel 13 -----------------------------
-
- all files:: ++jrb
- massive fixing of bugs, lots of re-arrangement etc.
-
- gdb.texinfo:: ++jrb
- newfile: an almost correct doc.
-
- ------------------------------- Patchlevel 14 -----------------------------
-
- m-atari.h:: ++jrb
- extract return value, copy less than 4 bytes from the correct
- place when type of return type != long.
-
- atarist.c:: ++jrb
- start child with Pexec modes 3+6
-
- CAUTION: CAUTION: CAUTION: CAUTION: CAUTION: CAUTION: CAUTION:
-
- TOS 1.4 required now, if you need to compile for older
- tos's, compile atarist.c with -DOLDTOS
-
- CAUTION: CAUTION: CAUTION: CAUTION: CAUTION: CAUTION: CAUTION:
-
- blockframe.c breakpoint.c dbxread.c eval.c findvar.c
- infcmd.c printcmd.c remote.c source.c st-core.c st-inflo.c
- st-infru.c stack.c symmisc.c symtab.c valarith.c
- valops.c valprint.c values.c:: ++jrb
- get rid of old init scheme. now all initialize functions
- are called initialize_<filename>() and they are called from
- init.c:initialize_all_files() (which in turn is called from main).
-
- init.c:: ++jrb
- new file, call all the initialize_<filename> functions.
-
- firstfile.c, lastfile.c, initialize.h:: ++jrb
- not used anymore
-
- Makefile:: ++jrb
- add target for init.o. delete firstfile.* lastfile.* initialize.*
-
- m-atari.h:: ++jrb
- had forgotten to revert back #def of FRAME_CHAIN, as a result some
- frames were setup incorrectly.
-
- atarist.c:: andreas
- (st_child_read_word, st_child_write_word): check for vaild address
- within child memory.
-
- source.c:: andreas
- (print_source_lines): allow last line of file to be printed.
-
- gnu-out.h:: ++jrb
- sync it up with util/gnu-out.h
-
- dbxread.c:: ++jrb
- get rid of the silly xxmalloc(), #define xxmalloc to xmalloc
-
- infcmd.c:: ++jrb
- - dont allow set-args after inferior has started. insist that
- args for inferior be specified either before issueing the initial
- run command with a set-arg or in the in the initial run command.
- - dont allow re-starting inferior.
-
- atarist.c:: ++jrb
- dont allow inferior to be re-run ever. returns special pid == -9999
- after issueing error. does'nt exit right away as the user may want to
- examine stuff or issue other non-run commands before he actually quits
- GDB.
-
- atarist.c, main.c:: ++jrb
- more -D's give more debug info. also include version in debug.gdb
-
- dbxread.c:: ++jrb
- - fix up bug in typevector allocation
- - dont free pending blocks (per gdb 4.?)
-
- main.c, util.c:: ++jrb
- save all vectors at startup, and restore them all back to the state
- at entry before exit. exit, now happend via routine really_exit().
-
- dbxread.c:: ++jrb
- - get rid of the silly symseg chain stuff that was cluttering up
- the code.
- - when reading pointer and func types dont do the smash_to_type
- this fixes the typedefs type * and just ordinary type *'s getting
- confused.
-
- symtab.c:: ++jrb
- - delete smash_to_* fns.
-
- all over:: ++jrb
- if '-L' option is specified, log all debugger output (append) to file
- "gdb.log". Use this with EXTREME CAUTION: it works with tos 1.4,
- but does'nt work with newer TOSs, and can potentially trash other
- files on the disk. YOU HAVE BEEN WARNED!
-
- ------------------------------- Patchlevel 15 -----------------------------
-
- atarist.c:: andreas
- (st_chile_read/write_word): addr can be one word below
- tpa start, it's on a 4 byte boundary; maybe it's too restrictive to
- test if addr is within child tpa, but it prevents bus errors if gdb
- tries to dereference uninitialized variables
-
- dbxread.c: andreas
- almost fool proof way to recognize -mshort; relies on the
- fact that `int' is always the first type defined (at least for GCC
- output, i don't know any other compiler for the st that generates
- dbx debugging info :-)
-
- expread.y: andreas
- always use 32 bits in expressions
-
- ------------------------------- Patchlevel 16 -----------------------------
-
- main.c:: ++jrb
- make gdb -L (logging) really work. We now buffer up the output,
- open the log file when buffer fills up, dump buffer, close file
- and so on.
- ------------------------------- Patchlevel 17 -----------------------------
-
- many places:: ++jrb
- undo all changes since PL 17, because nothing was working. Sigh!
-
- atarist.c:: andreas
- (st_execle_kludge): convert PATH as in spawn..;
- don't do anything but longjmp after Pexec because the stack is
- *very* small!
- (re_comp): free compiled regex
- (st_child_read_word, st_child_write_word): handle bus errors
- (kill): allow GDB output to be interrupted
-
- utils.c:: andreas
- (quit): handle interrupts
-
- dbxread.c:: andreas
- handle gcc-2 symbol table extensions
-
- findvar.c:: andreas
- correctly handle function types
-
- infcmd.c:: andreas
- (run_command): avoid confusion after user tried to
- start program again
- (finish_command): correctly handle function return type
-
- main.c:: andreas
- (Save_all_vectors): Setexc(N,-1) does work
-
- source.c:: andreas
- (directory_command): correctly handle atari-like
- absolute file names
-
- ------------------------------- Patchlevel 18 -----------------------------
-