Mercurial > hg > xemacs-beta
annotate etc/DEBUG @ 1315:70921960b980
[xemacs-hg @ 2003-02-20 08:19:28 by ben]
check in makefile fixes et al
Makefile.in.in: Major surgery. Move all stuff related to building anything in the
src/ directory into src/. Simplify the dependencies -- everything
in src/ is dependent on the single entry `src' in MAKE_SUBDIRS.
Remove weirdo targets like `all-elc[s]', dump-elc[s], etc.
mule/mule-msw-init.el: Removed.
Delete this file.
mule/mule-win32-init.el: New file, with stuff from mule-msw-init.el -- not just for MS Windows
native, boys and girls!
bytecomp.el: Change code inserted to catch trying to load a Mule-only .elc
file in a non-Mule XEmacs. Formerly you got the rather cryptic
"The required feature `mule' cannot be provided". Now you get
"Loading this file requires Mule support".
finder.el: Remove dependency on which directory this function is invoked
from.
update-elc.el: Don't mess around with ../src/BYTECOMPILE_CHANGE. Now that
Makefile.in.in and xemacs.mak are in sync, both of them use
NEEDTODUMP and the other one isn't used.
dumped-lisp.el: Rewrite in terms of `list' and `nconc' instead of assemble-list, so
we can have arbitrary forms, not just `when-feature'.
very-early-lisp.el: Nuke this file.
finder-inf.el, packages.el, update-elc.el, update-elc-2.el, loadup.el, make-docfile.el: Eliminate references to very-early-lisp.
msw-glyphs.el: Comment clarification.
xemacs.mak: Add macros DO_TEMACS, DO_XEMACS, and a few others; this macro
section is now completely in sync with src/Makefile.in.in. Copy
check-features, load-shadows, and rebuilding finder-inf.el from
src/Makefile.in.in. The main build/dump/recompile process is now
synchronized with src/Makefile.in.in. Change `WARNING' to `NOTE'
and `error checking' to `error-checking' TO avoid tripping
faux warnings and errors in the VC++ IDE.
Makefile.in.in: Major surgery. Move all stuff related to building anything in the
src/ directory from top-level Makefile.in.in to here. Simplify
the dependencies. Rearrange into logical subsections.
Synchronize the main compile/dump/build-elcs section with
xemacs.mak, which is already clean and in good working order.
Remove weirdo targets like `all-elc[s]', dump-elc[s], etc. Add
additional levels of macros \(e.g. DO_TEMACS, DO_XEMACS,
TEMACS_BATCH, XEMACS_BATCH, XEMACS_BATCH_PACKAGES) to factor out
duplicated stuff. Clean up handling of "HEAP_IN_DATA" (Cygwin) so
it doesn't need to ignore the return value from dumping. Add
.NO_PARALLEL since various aspects of building and dumping must be
serialized but do not always have dependencies between them
(this is impossible in some cases). Everything related to src/
now gets built in one pass in this directory by just running
`make' (except the Makefiles themselves and config.h, paths.h,
Emacs.ad.h, and other generated .h files).
console.c: Update list of possibly valid console types.
emacs.c: Rationalize the specifying and handling of the type of the first
frame. This was originally prompted by a workspace in which I got
GTK to compile under C++ and in the process fixed it so it could
coexist with X in the same build -- hence, a combined
TTY/X/MS-Windows/GTK build is now possible under Cygwin. (However,
you can't simultaneously *display* more than one kind of device
connection -- but getting that to work is not that difficult.
Perhaps a project for a bored grad student. I (ben) would do it
but don't see the use.) To make sense of this, I added new
switches that can be used to specifically indicate the window
system: -x [aka --use-x], -tty \[aka --use-tty], -msw [aka
--use-ms-windows], -gtk [aka --use-gtk], and -gnome [aka
--use-gnome, same as --use-gtk]. -nw continues as an alias for
-tty. When none have been given, XEmacs checks for other
parameters implying particular device types (-t -> tty, -display
-> x [or should it have same treatment as DISPLAY below?]), and
has ad-hoc logic afterwards: if env var DISPLAY is set, use x (or
gtk? perhaps should check whether gnome is running), else MS
Windows if it exsits, else TTY if it exists, else stream, and you
must be running in batch mode. This also fixes an existing bug
whereby compiling with no x, no mswin, no tty, when running non-
interactively (e.g. to dump) I get "sorry, must have TTY support".
emacs.c: Turn on Vstack_trace_on_error so that errors are debuggable even
when occurring extremely early in reinitialization.
emacs.c: Try to make sure that the user can see message output under
Windows (i.e. it doesn't just disappear right away) regardless of
when it occurs, e.g. in the middle of creating the first frame.
emacs.c: Define new function `emacs-run-status', indicating whether XEmacs
is noninteractive or interactive, whether raw,
post-dump/pdump-load or run-temacs, whether we are dumping,
whether pdump is in effect.
event-stream.c: It's "mommas are fat", not "momas are fat".
Fix other typo.
event-stream.c: Conditionalize in_menu_callback check on HAVE_MENUBARS,
because it won't exist on w/o menubar support,
lisp.h: More hackery on RETURN_NOT_REACHED. Cygwin v3.2 DOES complain here
if RETURN_NOT_REACHED() is blank, as it is for GCC 2.5+. So make it
blank only for GCC 2.5 through 2.999999999999999.
Declare Vstack_trace_on_error.
profile.c: Need to include "profile.h" to fix warnings.
sheap.c: Don't fatal() when need to rerun Make, just stderr_out() and exit(0).
That way we can distinguish between a dumping failing expectedly
(due to lack of stack space, triggering another dump) and unexpectedly,
in which case, we want to stop building. (or go on, if -K is given)
syntax.c, syntax.h: Use ints where they belong, and enum syntaxcode's where they belong,
and fix warnings thereby.
syntax.h: Fix crash caused by an edge condition in the syntax-cache macros.
text.h: Spacing fixes.
xmotif.h: New file, to get around shadowing warnings.
EmacsManager.c, event-Xt.c, glyphs-x.c, gui-x.c, input-method-motif.c, xmmanagerp.h, xmprimitivep.h: Include xmotif.h.
alloc.c: Conditionalize in_malloc on ERROR_CHECK_MALLOC.
config.h.in, file-coding.h, fileio.c, getloadavg.c, select-x.c, signal.c, sysdep.c, sysfile.h, systime.h, text.c, unicode.c: Eliminate HAVE_WIN32_CODING_SYSTEMS, use WIN32_ANY instead.
Replace defined (WIN32_NATIVE) || defined (CYGWIN) with WIN32_ANY.
lisp.h: More futile attempts to walk and chew gum at the same time when
dealing with subr's that don't return.
| author | ben |
|---|---|
| date | Thu, 20 Feb 2003 08:19:44 +0000 |
| parents | 376386a54a3c |
| children |
| rev | line source |
|---|---|
| 0 | 1 Debugging GNU Emacs |
| 2 Copyright (c) 1985 Richard M. Stallman. | |
| 3 | |
| 4 Permission is granted to anyone to make or distribute verbatim copies | |
| 5 of this document as received, in any medium, provided that the | |
| 6 copyright notice and permission notice are preserved, | |
| 7 and that the distributor grants the recipient permission | |
| 8 for further redistribution as permitted by this notice. | |
| 9 | |
| 10 Permission is granted to distribute modified versions | |
| 11 of this document, or of portions of it, | |
| 12 under the above conditions, provided also that they | |
| 13 carry prominent notices stating who last changed them. | |
| 14 | |
| 15 On 4.2 you will probably find that dbx does not work for | |
| 16 debugging GNU Emacs. For one thing, dbx does not keep the | |
| 17 inferior process's terminal modes separate from its own. | |
| 18 For another, dbx does not put the inferior in a separate | |
| 19 process group, which makes trouble when an inferior uses | |
| 20 interrupt input, which GNU Emacs must do on 4.2. | |
| 21 | |
| 22 dbx has also been observed to have other problems, | |
| 23 such as getting incorrect values for register variables | |
| 24 in stack frames other than the innermost one. | |
| 25 | |
| 26 The Emacs distribution now contains GDB, the new source-level | |
| 27 debugger for the GNU system. GDB works for debugging Emacs. | |
| 28 GDB currently runs on vaxes under 4.2 and on Sun 2 and Sun 3 | |
| 29 systems. | |
| 30 | |
| 31 | |
| 32 ** Some useful techniques | |
| 33 | |
| 34 `Fsignal' is a very useful place to stop in. | |
| 35 All Lisp errors go through there. | |
| 36 | |
| 37 It is useful, when debugging, to have a guaranteed way | |
| 38 to return to the debugger at any time. If you are using | |
| 39 interrupt-driven input, which is the default, then Emacs is using | |
| 40 RAW mode and the only way you can do it is to store | |
| 41 the code for some character into the variable stop_character: | |
| 42 | |
| 43 set stop_character = 29 | |
| 44 | |
| 45 makes Control-] (decimal code 29) the stop character. | |
| 46 Typing Control-] will cause immediate stop. You cannot | |
| 47 use the set command until the inferior process has been started. | |
| 48 Put a breakpoint early in `main', or suspend the Emacs, | |
| 49 to get an opportunity to do the set command. | |
| 50 | |
| 51 If you are using cbreak input (see the Lisp function set-input-mode), | |
| 52 then typing Control-g will cause a SIGINT, which will return control | |
| 53 to the debugger immediately unless you have done | |
| 54 | |
| 55 ignore 3 (in dbx) | |
| 56 or handle 3 nostop noprint (in gdb) | |
| 57 | |
| 58 You will note that most of GNU Emacs is written to avoid | |
| 59 declaring a local variable in an inner block, even in | |
| 60 cases where using one would be the cleanest thing to do. | |
| 61 This is because dbx cannot access any of the variables | |
| 62 in a function which has even one variable defined in an | |
| 63 inner block. A few functions in GNU Emacs do have variables | |
| 64 in inner blocks, only because I wrote them before realizing | |
| 65 that dbx had this problem and never rewrote them to avoid it. | |
| 66 | |
| 67 I believe that GDB does not have such a problem. | |
| 68 | |
| 69 | |
| 70 ** Examining Lisp object values. | |
| 71 | |
| 72 When you have a live process to debug, and it has not encountered a | |
| 73 fatal error, you can use the GDB command `pr'. First print the value | |
| 74 in the ordinary way, with the `p' command. Then type `pr' with no | |
| 75 arguments. This calls a subroutine which uses the Lisp printer. | |
| 76 | |
| 77 If you can't use this command, either because the process can't run | |
| 78 a subroutine or because the data is invalid, you can fall back on | |
| 79 lower-level commands. | |
| 80 | |
| 81 Use the `xtype' command to print out the data type of the last data | |
| 82 value. Once you know the data type, use the command that corresponds | |
| 83 to that type. Here are these commands: | |
| 84 | |
| 85 xint xptr xwindow xmarker xoverlay xmiscfree xintfwd xboolfwd xobjfwd | |
| 86 xbufobjfwd xkbobjfwd xbuflocal xbuffer xsymbol xstring xvector xframe | |
| 87 xwinconfig xcompiled xcons xcar xcdr xsubr xprocess xfloat xscrollbar | |
| 88 | |
| 89 Each one of them applies to a certain type or class of types. | |
| 90 (Some of these types are not visible in Lisp, because they exist only | |
| 91 internally.) | |
| 92 | |
| 93 Each x... command prints some information about the value, and | |
| 94 produces a GDB value (subsequently available in $) through which you | |
| 95 can get at the rest of the contents. | |
| 96 | |
| 97 In general, most of the rest of the contents will be addition Lisp | |
| 98 objects which you can examine in turn with the x... commands. | |
| 99 | |
| 100 ** If GDB does not run and your debuggers can't load Emacs. | |
| 101 | |
| 102 On some systems, no debugger can load Emacs with a symbol table, | |
| 103 perhaps because they all have fixed limits on the number of symbols | |
| 104 and Emacs exceeds the limits. Here is a method that can be used | |
| 105 in such an extremity. Do | |
| 106 | |
| 107 nm -n temacs > nmout | |
| 108 strip temacs | |
| 109 adb temacs | |
| 110 0xd:i | |
| 111 0xe:i | |
| 112 14:i | |
| 113 17:i | |
| 114 :r -l loadup (or whatever) | |
| 115 | |
| 116 It is necessary to refer to the file `nmout' to convert | |
| 117 numeric addresses into symbols and vice versa. | |
| 118 | |
| 119 It is useful to be running under a window system. | |
| 120 Then, if Emacs becomes hopelessly wedged, you can create | |
| 121 another window to do kill -9 in. kill -ILL is often | |
| 122 useful too, since that may make Emacs dump core or return | |
| 123 to adb. | |
| 124 | |
| 125 | |
| 126 ** Debugging incorrect screen updating. | |
| 127 | |
| 128 To debug Emacs problems that update the screen wrong, it is useful | |
| 129 to have a record of what input you typed and what Emacs sent to the | |
| 130 screen. To make these records, do | |
| 131 | |
| 132 (open-dribble-file "~/.dribble") | |
| 133 (open-termscript "~/.termscript") | |
| 134 | |
| 135 The dribble file contains all characters read by Emacs from the | |
| 136 terminal, and the termscript file contains all characters it sent to | |
| 137 the terminal. The use of the directory `~/' prevents interference | |
| 138 with any other user. | |
| 139 | |
| 140 If you have irreproducible display problems, put those two expressions | |
| 141 in your ~/.emacs file. When the problem happens, exit the Emacs that | |
| 142 you were running, kill it, and rename the two files. Then you can start | |
| 143 another Emacs without clobbering those files, and use it to examine them. |