I guess this ends the hunt for what change fixed 24.3 for msvc.
I spent several hours trying to track it down, but couldn't. My closest guest
is among the vector/list reading code. Lots of changes happened in this area.
Nothing very obvious anyway.

It is very unlucky, because I compile the trunk quite often and this bug never surfaced
for quite a long time (pre 24.1).

I guess the best option is to use the trunk as per Eli's proposal.

Fabrice


2013/3/15 李丁 <iamliding@gmail.com>
Though there are some minor errors (easy to fix), the latest trunk compiles fine with msvc.

2013/3/15 Fabrice Popineau <fabrice.popineau@gmail.com>
李丁:  could you try to compile the trunk with msvc and confirm that it is working for you?
I would be very glad to hear a positive report, meaning I didn't mess things up.

Fabrice


2013/3/15 李丁 <iamliding@gmail.com>
李丁, do you also compile Emacs 24.3 as a 64-bit executable?  Or do
you build it as a 32-bit executable?

I compiled Emacs as a 32-bit executable. 

If I run only temacs.exe without any arguments, I get a very quick backtrace:

I also traced the execution of `temacs -batch -l loadup dump' with windbg, and got a similar stack trace as Fabrice (I breakpointed Fgarbage_collect):

temacs!Fgarbage_collect [d:\data\projects\emacs-24.3\src\alloc.c @ 5094]
temacs!maybe_gc+0x3e [d:\data\projects\emacs-24.3\src\lisp.h @ 3717]
temacs!eval_sub+0xda [d:\data\projects\emacs-24.3\src\eval.c @ 2042]
temacs!readevalloop+0x600 [d:\data\projects\emacs-24.3\src\lread.c @ 1843]
temacs!Fload+0xb86 [d:\data\projects\emacs-24.3\src\lread.c @ 1317]
temacs!eval_sub+0x5da [d:\data\projects\emacs-24.3\src\eval.c @ 2159]
temacs!Feval+0x60 [d:\data\projects\emacs-24.3\src\eval.c @ 2005]
temacs!top_level_2+0x15 [d:\data\projects\emacs-24.3\src\keyboard.c @ 1177]
temacs!internal_condition_case+0xde [d:\data\projects\emacs-24.3\src\eval.c @ 1289]
temacs!top_level_1+0x26 [d:\data\projects\emacs-24.3\src\keyboard.c @ 1185]
temacs!internal_catch+0x97 [d:\data\projects\emacs-24.3\src\eval.c @ 1060]
temacs!command_loop+0x69 [d:\data\projects\emacs-24.3\src\keyboard.c @ 1146]
temacs!recursive_edit_1+0x71 [d:\data\projects\emacs-24.3\src\keyboard.c @ 779]
temacs!Frecursive_edit+0x101 [d:\data\projects\emacs-24.3\src\keyboard.c @ 844]
temacs!main+0xae7 [d:\data\projects\emacs-24.3\src\emacs.c @ 1530]
temacs!__tmainCRTStartup+0x1bf [f:\dd\vctools\crt_bld\self_x86\crt\src\crtexe.c @ 555]
temacs!mainCRTStartup+0xf [f:\dd\vctools\crt_bld\self_x86\crt\src\crtexe.c @ 371]
temacs!_start+0x62 [d:\data\projects\emacs-24.3\src\unexw32.c @ 134]

When I stepped out Fgarbage_collect, the error occurred just after this gc at here in eval_sub:

if (!CONSP (fun))
xsignal1 (Qinvalid_function, original_fun);

I also tried removing most of loadup.el, and temacs can execute only the first few lines, and even an additional `(+ 1 1)' caused temacs to exit with the "DEAD" message. As I previously reported, I guess that the first garbage collection cycle does not mark any read-in form, thus all of them are collected (and their car set to Vdead).


2013/3/15 Eli Zaretskii <eliz@gnu.org>
> From: Fabrice Popineau <fabrice.popineau@gmail.com>
> Date: Thu, 14 Mar 2013 20:28:45 +0100
> Cc: 李丁 <iamliding@gmail.com>, 13939 <13939@debbugs.gnu.org>
>
> I tried importing this change in editfns.c but it doesn't change anything.
>
> If I run only temacs.exe without any arguments, I get a very quick
> backtrace:
>
> C:\>"C:\Source\XEmTeX\emacs\emacs-24.3\src/obj-spd/AMD64/temacs.exe"
> Loading loadup.el (source)...
> Invalid function: "DEAD"
>
> > temacs.exe!eval_sub(__int64 form) Line 2195 C
>   temacs.exe!readevalloop(__int64 readcharfun, _iobuf * stream, __int64
> sourcename, char printflag, __int64 unibyte, __int64 readfun, __int64
> start, __int64 end) Line 1845 C

Thanks.

李丁, do you also compile Emacs 24.3 as a 64-bit executable?  Or do
you build it as a 32-bit executable?