* Compilation problems with latest MSVC
@ 2006-12-24 11:25 Eli Zaretskii
2006-12-24 11:35 ` Lennart Borgman
2006-12-29 2:40 ` David Robinow
0 siblings, 2 replies; 14+ messages in thread
From: Eli Zaretskii @ 2006-12-24 11:25 UTC (permalink / raw)
It seems like latest versions of MSVC are incompatible with the Emacs
build procedure (MSVC 6 compiles the current CVS just fine). See the
error messages below; this is with Visual Studio 2005 Express Edition,
the MS ``freeware'' distribution.
I could try working on these when I have time, but the question is: do
we care? Is it enough to say that the last version of MSVC we support
is v6? (Does anyone know if Studio 7 is okay?)
>nmake bootstrap
Microsoft (R) Program Maintenance Utility Version 7.00.8882
Copyright (C) Microsoft Corp 1988-2000. All rights reserved.
cl -I. -DWIN32_LEAN_AND_MEAN -D_WIN32_WINNT=0x0400 -nologo -D_X86_=1 -c -Zel -W2 -H63 -Oxsb2 -Oy- -G6dF -Zp8 -Zi -Di386 -D_CRTAPI1=_cdecl -Foobj-spd/i386\ addsection.c
cl : Command line warning D9035 : option 'Ze' has been deprecated and will be removed in a future release
cl : Command line warning D9035 : option 'H' has been deprecated and will be removed in a future release
cl : Command line warning D9002 : ignoring unknown option '-G6'
addsection.c
link -out:obj-spd/i386/addsection.exe -nologo -release -incremental:no -version:3.10 -swaprun:cd -swaprun:net setargv.obj -debug:full -debugtype:both obj-spd/i386/addsection.obj libc.lib oldnames.lib user32.lib
LINK : warning LNK4224: /DEBUG:FULL is no longer supported; ignored
LINK : fatal error LNK1117: syntax error in option 'debugtype:both'
NMAKE : fatal error U1077: 'link' : return code '0x45d'
Stop.
The compiler identifies itself as follows:
Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: Compilation problems with latest MSVC
2006-12-24 11:25 Compilation problems with latest MSVC Eli Zaretskii
@ 2006-12-24 11:35 ` Lennart Borgman
2006-12-29 2:40 ` David Robinow
1 sibling, 0 replies; 14+ messages in thread
From: Lennart Borgman @ 2006-12-24 11:35 UTC (permalink / raw)
Cc: emacs-devel
Eli Zaretskii wrote:
> It seems like latest versions of MSVC are incompatible with the Emacs
> build procedure (MSVC 6 compiles the current CVS just fine). See the
> error messages below; this is with Visual Studio 2005 Express Edition,
> the MS ``freeware'' distribution.
>
> I could try working on these when I have time, but the question is: do
> we care? Is it enough to say that the last version of MSVC we support
> is v6? (Does anyone know if Studio 7 is okay?)
I think it would be good to be able to compile with MSVC 7 too. I
sometimes wonder if strange problems I see have something to do with
bugs in the gcc compiler.
I am not saying that there are any problems with the gcc compiler. The
problems I see are stability problems on my pc. I hoped they would be
gone with my new pc using xp, but I have seen some problems again. Of
course I have no idea which program causes this, but it would be fine to
sometimes be able to run Emacs compiled with MSVC and see if that makes
any difference. (The instabilities are very rare however so it is very
hard to find out what is wrong. Probably some resource problem.)
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: Compilation problems with latest MSVC
2006-12-24 11:25 Compilation problems with latest MSVC Eli Zaretskii
2006-12-24 11:35 ` Lennart Borgman
@ 2006-12-29 2:40 ` David Robinow
2006-12-29 3:01 ` David Robinow
2006-12-29 16:29 ` Jason Rumney
1 sibling, 2 replies; 14+ messages in thread
From: David Robinow @ 2006-12-29 2:40 UTC (permalink / raw)
Cc: emacs-devel
On 12/24/06, Eli Zaretskii <eliz@gnu.org> wrote:
> It seems like latest versions of MSVC are incompatible with the Emacs
> build procedure (MSVC 6 compiles the current CVS just fine). See the
> error messages below; this is with Visual Studio 2005 Express Edition,
> the MS ``freeware'' distribution.
>
> I could try working on these when I have time, but the question is: do
> we care? Is it enough to say that the last version of MSVC we support
> is v6? (Does anyone know if Studio 7 is okay?)
Both Visual Studio .NET and Visual Studio .NET 2003 work fine. I've
been compiling Emacs with them both since they came out.
There is a warning that debugtype:both is no longer supported. I
don't remember which type it actually uses. As shown below, this
warning became an error in Visual Studio 2005.
>
> >nmake bootstrap
>
> Microsoft (R) Program Maintenance Utility Version 7.00.8882
> Copyright (C) Microsoft Corp 1988-2000. All rights reserved.
>
> cl -I. -DWIN32_LEAN_AND_MEAN -D_WIN32_WINNT=0x0400 -nologo -D_X86_=1 -c -Zel -W2 -H63 -Oxsb2 -Oy- -G6dF -Zp8 -Zi -Di386 -D_CRTAPI1=_cdecl -Foobj-spd/i386\ addsection.c
> cl : Command line warning D9035 : option 'Ze' has been deprecated and will be removed in a future release
> cl : Command line warning D9035 : option 'H' has been deprecated and will be removed in a future release
> cl : Command line warning D9002 : ignoring unknown option '-G6'
> addsection.c
> link -out:obj-spd/i386/addsection.exe -nologo -release -incremental:no -version:3.10 -swaprun:cd -swaprun:net setargv.obj -debug:full -debugtype:both obj-spd/i386/addsection.obj libc.lib oldnames.lib user32.lib
> LINK : warning LNK4224: /DEBUG:FULL is no longer supported; ignored
> LINK : fatal error LNK1117: syntax error in option 'debugtype:both'
> NMAKE : fatal error U1077: 'link' : return code '0x45d'
> Stop.
>
> The compiler identifies itself as follows:
>
> Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86
Visual Studio 6: Version 12.00.8168 MSC_VER == 1200 Linker
version 6.00.8477
NMAKE_VER = 6.00.8168.0
Visual Studio .NET: Version 13.00.9466 Linker version 7.00.9466
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: Compilation problems with latest MSVC
2006-12-29 2:40 ` David Robinow
@ 2006-12-29 3:01 ` David Robinow
2006-12-29 11:40 ` Eli Zaretskii
2006-12-29 16:29 ` Jason Rumney
1 sibling, 1 reply; 14+ messages in thread
From: David Robinow @ 2006-12-29 3:01 UTC (permalink / raw)
Cc: emacs-devel
On 12/28/06, David Robinow <drobinow@gmail.com> wrote:
> On 12/24/06, Eli Zaretskii <eliz@gnu.org> wrote:
> > It seems like latest versions of MSVC are incompatible with the Emacs
> > ....
Sorry. Fumble-fingers interrupted my previous post.
Below are values for distributed versions of Visual Studio. It seems
some hacking on nmake.defs and makefile.w32-in should be able to
eliminate the warnings.
Visual Studio 6: CL Version 12.00.8168 MSC_VER = 1200
Linker version 6.00.8477 _NMAKE_VER = 6.00.8168.0
Visual Studio .NET: CL Version 13.00.9466 MSC_VER = 1300
Linker version 7.00.9466 _NMAKE_VER = 7.0.9466
Visual Studio .NET 2003: CL Version 13.10.3077 MSC_VER = 1310
Linker vers. 7.10.3077 _NMAKE_VER = 7.10.3077
I tried to get the free-beer 2005 compiler to work about a month ago.
I edited out the offending debugtype:both and got quite a bit
farther. However, I had some other problem (which I can't remember
now) and ran out of time to play with it.
I'll try to report back after the first of the year.
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: Compilation problems with latest MSVC
2006-12-29 3:01 ` David Robinow
@ 2006-12-29 11:40 ` Eli Zaretskii
2006-12-29 16:25 ` Jason Rumney
0 siblings, 1 reply; 14+ messages in thread
From: Eli Zaretskii @ 2006-12-29 11:40 UTC (permalink / raw)
Cc: emacs-devel
> Date: Thu, 28 Dec 2006 22:01:31 -0500
> From: "David Robinow" <drobinow@gmail.com>
> Cc: emacs-devel@gnu.org
>
> On 12/28/06, David Robinow <drobinow@gmail.com> wrote:
> > On 12/24/06, Eli Zaretskii <eliz@gnu.org> wrote:
> > > It seems like latest versions of MSVC are incompatible with the Emacs
> > > ....
> Sorry. Fumble-fingers interrupted my previous post.
> Below are values for distributed versions of Visual Studio. It seems
> some hacking on nmake.defs and makefile.w32-in should be able to
> eliminate the warnings.
If you can suggest a safe way of making these work for both VS 6 and
the later versions, we could consider installing the changes before
the release. TIA
> Visual Studio 6: CL Version 12.00.8168 MSC_VER = 1200
> Linker version 6.00.8477 _NMAKE_VER = 6.00.8168.0
>
> Visual Studio .NET: CL Version 13.00.9466 MSC_VER = 1300
> Linker version 7.00.9466 _NMAKE_VER = 7.0.9466
>
> Visual Studio .NET 2003: CL Version 13.10.3077 MSC_VER = 1310
> Linker vers. 7.10.3077 _NMAKE_VER = 7.10.3077
Thanks. If nothing else, we should probably detect MSC_VER 1400 and
warn that it's unsupported.
> I tried to get the free-beer 2005 compiler to work about a month ago.
> I edited out the offending debugtype:both and got quite a bit
> farther. However, I had some other problem (which I can't remember
> now) and ran out of time to play with it.
I worked on that, and had quite a few more problems (related to the
changes in the system headers which come with VS 2005 and to the fact
that there's no single-threaded C library anymore). I solved those,
but then hit a brick wall when emacs-bootstrap built but crashed
during dumping. I don't have time to debug the crash, so unless
someone else figures out how to solve it, I guess we will just say
that VS 2005 is unsupported by Emacs 22.1.
I can post the patches I needed to get Emacs to compile with VS 2005,
if someone wants to work on this.
Btw, the bare-bones download of VS 2005 couldn't even begin compiling
the test programs during `configure', since there's no windows.h
header file. I needed to force VS 2005 to use the Platform SDK
include directory, to get it to compile. Maybe that's one reason
Emacs crashed, but then how to get the missing headers?
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: Compilation problems with latest MSVC
2006-12-29 11:40 ` Eli Zaretskii
@ 2006-12-29 16:25 ` Jason Rumney
2006-12-29 16:52 ` Eli Zaretskii
0 siblings, 1 reply; 14+ messages in thread
From: Jason Rumney @ 2006-12-29 16:25 UTC (permalink / raw)
Cc: emacs-devel, David Robinow
Eli Zaretskii wrote:
> I can post the patches I needed to get Emacs to compile with VS 2005,
> if someone wants to work on this.
>
I think I've figured out most of them, the remaining problem I'm having
is that strftime.c and editfns.c want to use _tzname, which is not
defined by any of the libraries.
The changes I had to make outside the makefiles seem harmless, so I've
checked them in.
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: Compilation problems with latest MSVC
2006-12-29 2:40 ` David Robinow
2006-12-29 3:01 ` David Robinow
@ 2006-12-29 16:29 ` Jason Rumney
1 sibling, 0 replies; 14+ messages in thread
From: Jason Rumney @ 2006-12-29 16:29 UTC (permalink / raw)
Cc: Eli Zaretskii, emacs-devel
David Robinow wrote:
> There is a warning that debugtype:both is no longer supported. I
> don't remember which type it actually uses. As shown below, this
> warning became an error in Visual Studio 2005.
I think it is OK to remove this option. It was useful when Emacs could
not be compiled with gcc on Windows, because it allowed people to debug
the precompiled binaries using gdb. But now that gcc is used to build
the precompiled binaries, we can assume that anyone compiling with msvc
will also have access to the microsoft debugger.
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: Compilation problems with latest MSVC
2006-12-29 16:25 ` Jason Rumney
@ 2006-12-29 16:52 ` Eli Zaretskii
2006-12-29 23:05 ` Jason Rumney
2006-12-29 23:43 ` Jason Rumney
0 siblings, 2 replies; 14+ messages in thread
From: Eli Zaretskii @ 2006-12-29 16:52 UTC (permalink / raw)
Cc: emacs-devel, drobinow
> Date: Fri, 29 Dec 2006 16:25:03 +0000
> From: Jason Rumney <jasonr@gnu.org>
> Cc: David Robinow <drobinow@gmail.com>, emacs-devel@gnu.org
>
> I think I've figured out most of them, the remaining problem I'm having
> is that strftime.c and editfns.c want to use _tzname, which is not
> defined by any of the libraries.
I think I solved that by having HAVE_TZNAME defined in src/s/ms-w32.h
only for versions of MSVC before 1400. If you think this is an okay
solution, I can dig out the patches (not that the above leaves any
place for doubt about the exact change).
> The changes I had to make outside the makefiles seem harmless, so I've
> checked them in.
Thanks.
How did you solve the problem with missing headers (windows.h and
probably the rest of Windows-specific ones)?
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: Compilation problems with latest MSVC
2006-12-29 16:52 ` Eli Zaretskii
@ 2006-12-29 23:05 ` Jason Rumney
2006-12-30 14:15 ` Eli Zaretskii
2006-12-29 23:43 ` Jason Rumney
1 sibling, 1 reply; 14+ messages in thread
From: Jason Rumney @ 2006-12-29 23:05 UTC (permalink / raw)
Cc: emacs-devel, drobinow
Eli Zaretskii wrote:
> How did you solve the problem with missing headers (windows.h and
> probably the rest of Windows-specific ones)?
>
I'm not using the no-cost version, so everything comes prepackaged. From
when I used the no-cost version in the past, there were several separate
downloads needed, and figuring out which ones and how to configure the
paths so the compiler can find everything was not an easy task, and
completely undocumented. It sounds like you are missing the Win32
Platform SDK.
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: Compilation problems with latest MSVC
2006-12-29 16:52 ` Eli Zaretskii
2006-12-29 23:05 ` Jason Rumney
@ 2006-12-29 23:43 ` Jason Rumney
2006-12-30 14:18 ` Eli Zaretskii
1 sibling, 1 reply; 14+ messages in thread
From: Jason Rumney @ 2006-12-29 23:43 UTC (permalink / raw)
Cc: emacs-devel, drobinow
Eli Zaretskii wrote:
> I think I solved that by having HAVE_TZNAME defined in src/s/ms-w32.h
> only for versions of MSVC before 1400. If you think this is an okay
> solution, I can dig out the patches (not that the above leaves any
> place for doubt about the exact change).
>
That was enough to get to the crash while dumping, but since tzname is
in the system headers, I think it is possible to find a better solution
if we can make it work.
The crash is happening within the initialization code before any Emacs
code gets run (__tmainCRTStartup). This code is calling _free_internal,
which we are redefining in gmalloc.c. I'm not sure if this can be fixed
without some major changes, so I'm inclined to put it aside until after
the release.
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: Compilation problems with latest MSVC
2006-12-29 23:05 ` Jason Rumney
@ 2006-12-30 14:15 ` Eli Zaretskii
0 siblings, 0 replies; 14+ messages in thread
From: Eli Zaretskii @ 2006-12-30 14:15 UTC (permalink / raw)
Cc: emacs-devel, drobinow
> Date: Fri, 29 Dec 2006 23:05:54 +0000
> From: Jason Rumney <jasonr@gnu.org>
> Cc: drobinow@gmail.com, emacs-devel@gnu.org
>
> Eli Zaretskii wrote:
> > How did you solve the problem with missing headers (windows.h and
> > probably the rest of Windows-specific ones)?
> >
> I'm not using the no-cost version, so everything comes prepackaged. From
> when I used the no-cost version in the past, there were several separate
> downloads needed, and figuring out which ones and how to configure the
> paths so the compiler can find everything was not an easy task, and
> completely undocumented. It sounds like you are missing the Win32
> Platform SDK.
No, I did install the Platform SDK (and tweaked the VS 2005 startup
script to invoke the SDK startup script, to set up INCLUDE and LIB
paths correctly). But I wasn't sure about the correct order of the
paths (since there could be duplicate libraries/headers). Also, the
Platform SDK had several versions since VS 2005 was released, and I
wasn't sure the last one that's available is compatible with VS 2005.
But since you've got the same crash with your prepackaged version, I
think my fears were unsubstantiated, and the crash is a real problem.
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: Compilation problems with latest MSVC
2006-12-29 23:43 ` Jason Rumney
@ 2006-12-30 14:18 ` Eli Zaretskii
2006-12-31 0:13 ` Jason Rumney
0 siblings, 1 reply; 14+ messages in thread
From: Eli Zaretskii @ 2006-12-30 14:18 UTC (permalink / raw)
Cc: emacs-devel, drobinow
> Date: Fri, 29 Dec 2006 23:43:17 +0000
> From: Jason Rumney <jasonr@gnu.org>
> Cc: drobinow@gmail.com, emacs-devel@gnu.org
>
> The crash is happening within the initialization code before any Emacs
> code gets run (__tmainCRTStartup). This code is calling _free_internal,
> which we are redefining in gmalloc.c.
That rings a bell: during linking, the linker complained about
multiple definitions of some memory-related functions, I think calloc
and realloc. Did you have those warnings in your build?
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: Compilation problems with latest MSVC
2006-12-30 14:18 ` Eli Zaretskii
@ 2006-12-31 0:13 ` Jason Rumney
2006-12-31 4:24 ` Eli Zaretskii
0 siblings, 1 reply; 14+ messages in thread
From: Jason Rumney @ 2006-12-31 0:13 UTC (permalink / raw)
Cc: emacs-devel, drobinow
Eli Zaretskii wrote:
>> Date: Fri, 29 Dec 2006 23:43:17 +0000
>> From: Jason Rumney <jasonr@gnu.org>
>> Cc: drobinow@gmail.com, emacs-devel@gnu.org
>>
>> The crash is happening within the initialization code before any Emacs
>> code gets run (__tmainCRTStartup). This code is calling _free_internal,
>> which we are redefining in gmalloc.c.
>>
>
> That rings a bell: during linking, the linker complained about
> multiple definitions of some memory-related functions, I think calloc
> and realloc. Did you have those warnings in your build?
>
Yes, they appear as errors if you link against libcmt.lib, unless you
include -force:multiple in LINK_FLAGS. I suspect that the single
threaded library is required for Emacs, as the multithreaded library
will be allocating heap memory for each thread behind your back to avoid
problems with global variables, so there will always be a conflict
between Emacs's and the C library's malloc routines.
I also tried linking against msvcrt.lib (the dynamic linked version of
libcmt). It doesn't complain about the multiple definitions, but there
are more unresolved symbols. I also recall Andrew Innes trying in the
past to get Emacs working with the dynamic linked c runtime, I'm not
sure if he succeeded in the end by ensuring that Emacs's memory
allocations do not intertwine with the system library's or whether he
gave up.
^ permalink raw reply [flat|nested] 14+ messages in thread
* Re: Compilation problems with latest MSVC
2006-12-31 0:13 ` Jason Rumney
@ 2006-12-31 4:24 ` Eli Zaretskii
0 siblings, 0 replies; 14+ messages in thread
From: Eli Zaretskii @ 2006-12-31 4:24 UTC (permalink / raw)
Cc: emacs-devel, drobinow
> Date: Sun, 31 Dec 2006 00:13:18 +0000
> From: Jason Rumney <jasonr@gnu.org>
> Cc: drobinow@gmail.com, emacs-devel@gnu.org
> >
> I suspect that the single threaded library is required for Emacs, as
> the multithreaded library will be allocating heap memory for each
> thread behind your back to avoid problems with global variables, so
> there will always be a conflict between Emacs's and the C library's
> malloc routines.
Alas, there's no single-threaded library that comes with VS 2005: they
removed it. The docs says "use the multithreaded library instead".
So it sounds like all later versions of MSVC will have this problem.
^ permalink raw reply [flat|nested] 14+ messages in thread
end of thread, other threads:[~2006-12-31 4:24 UTC | newest]
Thread overview: 14+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2006-12-24 11:25 Compilation problems with latest MSVC Eli Zaretskii
2006-12-24 11:35 ` Lennart Borgman
2006-12-29 2:40 ` David Robinow
2006-12-29 3:01 ` David Robinow
2006-12-29 11:40 ` Eli Zaretskii
2006-12-29 16:25 ` Jason Rumney
2006-12-29 16:52 ` Eli Zaretskii
2006-12-29 23:05 ` Jason Rumney
2006-12-30 14:15 ` Eli Zaretskii
2006-12-29 23:43 ` Jason Rumney
2006-12-30 14:18 ` Eli Zaretskii
2006-12-31 0:13 ` Jason Rumney
2006-12-31 4:24 ` Eli Zaretskii
2006-12-29 16:29 ` Jason Rumney
Code repositories for project(s) associated with this external index
https://git.savannah.gnu.org/cgit/emacs.git
https://git.savannah.gnu.org/cgit/emacs/org-mode.git
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.