This is GDB, the GNU source-level debugger, presently running under un*x. This is a pre-alpha version of GDB version 4, and has NOT been extensively tested. It surely has some bugs, both bugs that were present in version 3 and new bugs. I have filed all the bug reports and fixes mailed to bug-gdb, and the fixes in particular will move into these sources as I find the time. => THIS VERSION IS PARTICULARLY FRAGILE! <= It depends on a preliminary version of a new "binary file descriptor" library and a new global "include" directory, which are packaged separately from GDB. You must obtain, configure and build this library manually, then configure and build gdb. When building gdb's for multiple platforms, you must manually rebuild the bfd library separately for each platform. Yes, of course, we are working on this! FIXME! Configure bfd for your host system by: cd ../bfd edit Makefile make Then you can cd ../gdb-whatever, and config and build gdb. This release moves the generic GNU include files, the BFD library, and the getopt routines into the parent directory of gdb. The idea is that a variety of GNU tools can share a common copy of these things. A summary of features new since gdb-3.5 is in the file `WHATS.NEW'. The best way to build GDB, in my opinion, is in a subdirectory. I use a naming convention "=XXX" where XXX is the machine type I'm building for. Nothing depends on this, it's just how I remember which subdirectories are what. So, once you have the BFD library built for that machine, you can do: cd gdb-x.yy (the directory where this README is) mkdir =XXX (e.g. mkdir =vax) cd =XXX ../config.gdb machine make Machine is like "vax" or "sun4". For more information type `../config.gdb'. Once you have done that, just `make' will do everything, producing an executable `gdb' in this directory. You can also build gdb binaries in a completely different directory from its sources, by specifying "srcdir=YYY" to config.gdb, giving it an absolute or relative path to the source directory. GDB can be used as a cross-debugger, running on a machine of one type while debugging a program running on a machine of another type. You configure it this way by specifying `config.gdb host target' where host is where GDB runs, and target is where your program runs. If you want a new (current to this release) version of the manual, you will have to use the gdb.texinfo file provided with this distribution. For details see the texinfo manual (distributed with emacs and as a printed manual). About languages other than C... C++ support has been integrated into gdb. GDB should work with FORTRAN programs (if you have problem, please send a bug report; note that you may have to refer to some FORTRAN variables with a trailing underscore), but I am not aware of anyone who is working on getting it to use the syntax of any language other than C or C++. Pascal programs which use sets, subranges, file variables, or nested functions will not currently work. About kernel debugging... I have't done this myself so I can't really offer any advice. Remote debugging over serial lines is more like to be in a currently functioning state than the standalone gdb (kdb). FIXME. About remote debugging... [This section seems to be out of date, I have never seen the "rapp" program, though I would like to. FIXME.] `rapp' runs under unix and acts as a remote stub (like remote-multi.shar distributed with GDB version 3). Currently it just works over UDP (network), not over a serial line. To get it running * Compile GDB on the host machine as usual * Compile rapp on the target machine, giving for both host and target the type of the target machine * Install "gdb" in /etc/services on both machines. This will get reworked before the initial release of 4.x. FIXME. The two files remote-multi.shar and remote-sa.m68k.shar contain two examples of a remote stub to be used with remote.c. The the -multi file is a general stub that can probably be running on various different flavors of unix to allow debugging over a serial line from one machine to another. The remote-sa.m68k.shar is designed to run standalone on a 68k type cpu and communicate properley with the remote.c stub over a serial line. The files remote-eb.c and remote-nindy.c are two examples of remote interfaces for talking to existing ROM monitors (for the AMD 29000 and the Intel 960 repsectively). There is also a remote interface for the VxWorks realtime kernel, which communicates over TCP/IP, in remote-vx.c and the vx-share subdirectory. About reporting bugs... The correct address for reporting bugs found with gdb is "bug-gdb@prep.ai.mit.edu". Please email all bugs to that address. About xgdb... Hopefully a new xgdb will be in 4.x. xgdb.c was provided to us by the user community; it is not an integral part of the gdb distribution. The problem of providing visual debugging support on top of gdb is peripheral to the GNU project and (at least right now) we can't afford to put time into it. So while we will be happy to incorporate user fixes to xgdb.c, we do not guarantee that it will work and we will not fix bugs reported in it. See XGDB-README for one person's opinion about what is wrong with the current xgdb. Someone is working on writing a new XGDB, so improving (e.g. by fixing it so that it will work, if it doesn't currently) the current one is not worth it. For those intersted in auto display of source and the availability of an editor while debugging I suggest trying gdb-mode in gnu-emacs (Try typing M-x gdb RETURN). Comments on this mode are welcome. About the machine-dependent files... tconfig/ This contains Makefile stuff for when the target system is . It also specifies the name of the tm-XXX.h file for this machine. xconfig/ This contains Makefile stuff for when the host system is . It also specifies the name of the xm-XXX.h file for this machine. tm-XXX.h (tm.h is a link to this file, created by config.gdb). This file contains macro definitions that express information about the target machine's registers, stack frame format and instructions. xm-XXX.h (xm.h is a link to this file, created by config.gdb). This contains macro definitions describing the host system environment, such as byte order, host C compiler and library, ptrace support, and core file structure. -opcode.h -pinsn.c These files contain the information necessary to print instructions for your cpu type. -opcode.h includes some large initialized data structures, which is strange for a ".h" file, but it's OK since it is only included in one place. -opcode.h is shared between the debugger and the assembler (if the GNU assembler has been ported to that machine), whereas -pinsn.c is specific to GDB. -tdep.c This file contains any miscellaneous code required for this machine as a target. On some machines it doesn't exist at all. Its existence is specified in the tconfig/XXX file. -xdep.c This file contains any miscellaneous code required for this machine as a host. On some machines it doesn't exist at all. Its existence is specified in the xconfig/XXX file. infptrace.c This is the low level interface to inferior processes for systems using the Unix ptrace call in a vanilla way. Some systems have their own routines in -xdep.c. Whether or not it is used is specified in the xconfig/XXX file. coredep.c Machine and system-dependent aspects of reading core files. Some machines use coredep.c; some have the routines in -xdep.c. Whether or not it is used is specified in the xconfig/XXX file. Now that BFD is used to read core files, virtually all machines should use coredep.c and should just provide fetch_core_registers in -xdep.c. exec.c Machine and system-dependent aspects of reading executable files. Some machines use exec.c; some have the routines in -tdep.c Since BFD, virtually all machines should use exec.c. About writing code for GDB... We appreciate having users contribute code that is of general use, but for it to be included in future GDB releases it must be cleanly written. We do not want to include changes that will needlessly make future maintainance difficult. It is not much harder to do things right, and in the long term it is worth it to the GNU project, and probably to you individually as well. Please code according to the GNU coding standards. If you do not have a copy, you can request one by sending mail to gnu@prep.ai.mit.edu. If you make substantial changes, you'll have to file a copyright assignment with the Free Software Foundation before we can produce a release that includes your changes. Send mail requesting the copyright assignment to gnu@prep.ai.mit.edu. Do this early, like before the changes actually work, or even before you start them, because a manager or lawyer on your end will probably make this a slow process. Please try to avoid making machine-specific changes to machine-independent files. If this is unavoidable, put a hook in the machine-independent file which calls a (possibly) machine-dependent macro (for example, the IGNORE_SYMBOL macro can be used for any symbols which need to be ignored on a specific machine. Calling IGNORE_SYMBOL in dbxread.c is a lot cleaner than a maze of #if defined's). The machine-independent code should do whatever "most" machines want if the macro is not defined in param.h. Using #if defined can sometimes be OK (e.g. SET_STACK_LIMIT_HUGE) but should be conditionalized on a specific feature of an operating system (set in tm.h or xm.h) rather than something like #if defined(vax) or #if defined(SYSV). If you use an #ifdef on some symbol that is defined in a header file (e.g. #ifdef TIOCSETP), *please* make sure that you have #include'd the relevant header file in that module! It is better to replace entire routines which may be system-specific, rather than put in a whole bunch of hooks which are probably not going to be helpful for any purpose other than your changes. For example, if you want to modify dbxread.c to deal with DBX debugging symbols which are in COFF files rather than BSD a.out files, do something along the lines of a macro GET_NEXT_SYMBOL, which could have different definitions for COFF and a.out, rather than trying to put the necessary changes throughout all the code in dbxread.c that currently assumes BSD format. Please avoid duplicating code. For example, in GDB 3.x all the stuff in infptrace.c was duplicated in *-dep.c, and so changing something was very painful. Thus in GDB 4.x these have all been consolidated into infptrace.c. infptrace.c can deal with variations between systems the same way any system-independent file would (hooks, #if defined, etc.), and machines which are radically different don't need to use infptrace.c at all. The same was true of core_file_command and exec_file_command. About debugging gdb with itself... You probably want to do a "make TAGS" after you configure your distribution; this will put the machine dependent routines for your local machine where they will be accessed first by a M-period . Also, make sure that you've compiled gdb with your local cc or taken appropriate precautions regarding ansification of include files. See the Makefile for more information. When you run gdb in this directory, it will read a ".gdbinit" file that sets up some simple things to make debugging gdb easier. The "info" command, when executed without a subcommand in a gdb being debugged by gdb, will pop you back up to the top level gdb. See .gdbinit for details. (this is for editing this file with GNU emacs) Local Variables: mode: text End: