GNU bug report logs - #53377
28.0.91: Build fails on MSYS2/MinGW-w64 with Clang

Previous Next

To reply to this bug, email your comments to 53377 AT debbugs.gnu.org.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Óscar Fuentes <ofv <at> wanadoo.es>
To: bug-gnu-emacs <at> gnu.org
Subject: 28.0.91: Build fails on MSYS2/MinGW-w64 with Clang
Date: Thu, 20 Jan 2022 00:41:03 +0100

Building under the Clang environment (which uses clang instead of gcc)
fails with

$ make V=1
make -C lib all
make[1]: Entering directory '/d/dev/other/MINGW-packages/mingw-w64-emacs/src/build-CLANG64/lib'
clang -c -D__USE_MINGW_ANSI_STDIO=1   -MMD -MF deps/getopt.d -MP -Wno-switch -Wno-pointer-sign -Wno-string-plus-int -Wno-unknown-attributes -Wno-initializer-overrides -Wno-tautological-compare -Wno-tautological-constant-out-of-range-compare   -march=x86-64 -mtune=generic -O2 -pipe -I. -I../src -I../../emacs-28.0.91/lib -I../../emacs-28.0.91/lib/../src  -o getopt.o ../../emacs-28.0.91/lib/getopt.c
In file included from ../../emacs-28.0.91/lib/getopt.c:24:
In file included from ./getopt.h:52:
In file included from ./unistd.h:40:
In file included from C:/apps/msys64/clang64/x86_64-w64-mingw32/include/unistd.h:11:
In file included from ../../emacs-28.0.91/lib/../src/process.h:29:
In file included from ../../emacs-28.0.91/lib/../src/gnutls.h:31:
../../emacs-28.0.91/lib/../src/lisp.h:944:10: fatal error: 'globals.h' file not found
#include "globals.h"
         ^~~~~~~~~~~
1 error generated.
make[1]: *** [Makefile:84: getopt.o] Error 1
make[1]: Leaving directory '/d/dev/other/MINGW-packages/mingw-w64-emacs/src/build-CLANG64/lib'
make: *** [Makefile:436: lib] Error 2


globals.h does not exist. AFAIK it is a generated file, but on this
build it is not generated.

The configure invocation is

  configure \
    --prefix=/clang64 \
    --build=x86_64-w64-mingw32 \
    --with-modules \
    --without-dbus \
    --without-compress-install

Speaking of configure, it does not even try to locate several libraries.
The string "jpeg", for instance, only appears in config.log as

LIBJPEG=''

Message #8 received at 53377 <at> debbugs.gnu.org (full text, mbox):

From: Po Lu <luangruo <at> yahoo.com>
To: Óscar Fuentes <ofv <at> wanadoo.es>
Cc: 53377 <at> debbugs.gnu.org
Subject: Re: bug#53377: 28.0.91: Build fails on MSYS2/MinGW-w64 with Clang
Date: Thu, 20 Jan 2022 09:14:24 +0800

Óscar Fuentes <ofv <at> wanadoo.es> writes:

> Building under the Clang environment (which uses clang instead of gcc)
> fails with

Do we really want to support building Emacs with Clang on MS-Windows?

AFAIU, it's support for that platform is miles behind GCC's, so even if
this specific build error was fixed, there would be other problems
preventing Emacs from working there.

Message #11 received at 53377 <at> debbugs.gnu.org (full text, mbox):

From: Óscar Fuentes <ofv <at> wanadoo.es>
To: Po Lu <luangruo <at> yahoo.com>
Cc: 53377 <at> debbugs.gnu.org
Subject: Re: bug#53377: 28.0.91: Build fails on MSYS2/MinGW-w64 with Clang
Date: Thu, 20 Jan 2022 04:32:04 +0100

Po Lu <luangruo <at> yahoo.com> writes:

> Do we really want to support building Emacs with Clang on MS-Windows?
>
> AFAIU, it's support for that platform is miles behind GCC's,

Your information is outdated. The only clear advantage GCC has over
Clang is at being compatible with GCC.

> so even if this specific build error was fixed, there would be other
> problems preventing Emacs from working there.

Nobody is asking you to work on this bug report, so feel free to ignore
it and don't discourage others.

Message #14 received at 53377 <at> debbugs.gnu.org (full text, mbox):

From: Po Lu <luangruo <at> yahoo.com>
To: Óscar Fuentes <ofv <at> wanadoo.es>
Cc: 53377 <at> debbugs.gnu.org
Subject: Re: bug#53377: 28.0.91: Build fails on MSYS2/MinGW-w64 with Clang
Date: Thu, 20 Jan 2022 11:41:05 +0800

Óscar Fuentes <ofv <at> wanadoo.es> writes:

> Your information is outdated. The only clear advantage GCC has over
> Clang is at being compatible with GCC.

The last I tried, I could not even step through Clang-generated code
with GDB on MS-Windows.  The line numbers were all wrong.  Admittedly
that was 4 months ago, but I don't see that it's likely for much to have
changed since then.

So this is why I think we shouldn't worry about building on Windows with
Clang.  There are enough real problems to worry about.

Message #17 received at 53377 <at> debbugs.gnu.org (full text, mbox):

From: Óscar Fuentes <ofv <at> wanadoo.es>
To: Po Lu <luangruo <at> yahoo.com>
Cc: 53377 <at> debbugs.gnu.org
Subject: Re: bug#53377: 28.0.91: Build fails on MSYS2/MinGW-w64 with Clang
Date: Thu, 20 Jan 2022 04:59:55 +0100

Po Lu <luangruo <at> yahoo.com> writes:

> Óscar Fuentes <ofv <at> wanadoo.es> writes:
>
>> Your information is outdated. The only clear advantage GCC has over
>> Clang is at being compatible with GCC.
>
> The last I tried, I could not even step through Clang-generated code
> with GDB on MS-Windows.  The line numbers were all wrong.  Admittedly
> that was 4 months ago, but I don't see that it's likely for much to have
> changed since then.

Dunno, I use a debugger maybe once every several months, mostly on
GNU/Linux, where recent versions of Clang play reasonably well with gdb,
or at least not much worse than GCC. Besides, Clang has its own
debugger.

> So this is why I think we shouldn't worry about building on Windows with
> Clang.  There are enough real problems to worry about.

Then, by all means, keep working on those "real" problems and ignore
this issue. But please don't try to discourage others from working on
this.

Message #20 received at 53377 <at> debbugs.gnu.org (full text, mbox):

From: Po Lu <luangruo <at> yahoo.com>
To: Óscar Fuentes <ofv <at> wanadoo.es>
Cc: 53377 <at> debbugs.gnu.org
Subject: Re: bug#53377: 28.0.91: Build fails on MSYS2/MinGW-w64 with Clang
Date: Thu, 20 Jan 2022 12:52:45 +0800

Óscar Fuentes <ofv <at> wanadoo.es> writes:

> Then, by all means, keep working on those "real" problems and ignore
> this issue. But please don't try to discourage others from working on
> this.

It seems like a waste of time and effort, so I recommend that people not
work on it.  (Which is "discouragement", and for a valid reason.)

But keep in mind that nothing prevents any interested party from working
on it.

Message #23 received at 53377 <at> debbugs.gnu.org (full text, mbox):

From: Eli Zaretskii <eliz <at> gnu.org>
To: Óscar Fuentes <ofv <at> wanadoo.es>
Cc: 53377 <at> debbugs.gnu.org
Subject: Re: bug#53377: 28.0.91: Build fails on MSYS2/MinGW-w64 with Clang
Date: Thu, 20 Jan 2022 09:33:00 +0200

> From: Óscar Fuentes <ofv <at> wanadoo.es>
> Date: Thu, 20 Jan 2022 00:41:03 +0100
> 
> 
> Building under the Clang environment (which uses clang instead of gcc)
> fails with
> 
> $ make V=1
> make -C lib all
> make[1]: Entering directory '/d/dev/other/MINGW-packages/mingw-w64-emacs/src/build-CLANG64/lib'
> clang -c -D__USE_MINGW_ANSI_STDIO=1   -MMD -MF deps/getopt.d -MP -Wno-switch -Wno-pointer-sign -Wno-string-plus-int -Wno-unknown-attributes -Wno-initializer-overrides -Wno-tautological-compare -Wno-tautological-constant-out-of-range-compare   -march=x86-64 -mtune=generic -O2 -pipe -I. -I../src -I../../emacs-28.0.91/lib -I../../emacs-28.0.91/lib/../src  -o getopt.o ../../emacs-28.0.91/lib/getopt.c
> In file included from ../../emacs-28.0.91/lib/getopt.c:24:
> In file included from ./getopt.h:52:
> In file included from ./unistd.h:40:
> In file included from C:/apps/msys64/clang64/x86_64-w64-mingw32/include/unistd.h:11:
> In file included from ../../emacs-28.0.91/lib/../src/process.h:29:
> In file included from ../../emacs-28.0.91/lib/../src/gnutls.h:31:
> ../../emacs-28.0.91/lib/../src/lisp.h:944:10: fatal error: 'globals.h' file not found

Your problem is that lib/unistd.h is not supposed to exist in the
Windows builds: we don't produce it from lib/unistd.in.h.  So you need
to find out why in your case it was produced.  That's the immediate
reason for the failure.

I see another problem: MS-Windows builds are supposed to use the cpp
option "-I ../nt/inc", to pick up our replacements for missing and
incomplete stuff in system header files, and your compilation command
above doesn't include that flag.  This could be important in this case
because we have our own replacement for unistd.h in nt/inc/; the
system unistd.h is not supposed to be used at all in the Emacs build
on MS-Windows.  It is definitely important for the rest of the build.

So it sounds like you are not using the build machinery for compiling
Emacs on MS-Windows, and if that is intentional, you are up for a
bumpy ride, and reporting bugs about this adventure is not necessarily
a good idea.  You should first make your build process follow the path
of MinGW, otherwise many things will break, either vociferously or
(worse) silently.

> Speaking of configure, it does not even try to locate several libraries.
> The string "jpeg", for instance, only appears in config.log as
> 
> LIBJPEG=''

So something else is seriously wrong, because on my system I have
this:

  configure:18511: checking for jpeglib 6b or later
  configure:18555: result: -ljpeg
  [...]
  emacs_cv_jpeglib=-ljpeg
  [...]
  #define HAVE_JPEG 1
  [...]
  #define EMACS_CONFIG_FEATURES "ACL GIF GMP GNUTLS HARFBUZZ JPEG JSON LCMS2 LIBXML2 MODULES NOTIFY W32NOTIFY PDUMPER PNG RSVG SOUND SQLITE3 THREADS TIFF TOOLKIT_SCROLL_BARS WEBP XPM ZLIB"

Message #26 received at 53377 <at> debbugs.gnu.org (full text, mbox):

From: Eli Zaretskii <eliz <at> gnu.org>
To: Po Lu <luangruo <at> yahoo.com>
Cc: ofv <at> wanadoo.es, 53377 <at> debbugs.gnu.org
Subject: Re: bug#53377: 28.0.91: Build fails on MSYS2/MinGW-w64 with Clang
Date: Thu, 20 Jan 2022 09:40:04 +0200

> Resent-From: Po Lu <luangruo <at> yahoo.com>
> Original-Sender: "Debbugs-submit" <debbugs-submit-bounces <at> debbugs.gnu.org>
> Resent-CC: bug-gnu-emacs <at> gnu.org
> Resent-Sender: help-debbugs <at> gnu.org
> Cc: 53377 <at> debbugs.gnu.org
> Date: Thu, 20 Jan 2022 09:14:24 +0800
> From:  Po Lu via "Bug reports for GNU Emacs,
>  the Swiss army knife of text editors" <bug-gnu-emacs <at> gnu.org>
> 
> Óscar Fuentes <ofv <at> wanadoo.es> writes:
> 
> > Building under the Clang environment (which uses clang instead of gcc)
> > fails with
> 
> Do we really want to support building Emacs with Clang on MS-Windows?

It depends on what would such support require.  Judging by what I see
in the original report, we are very far from understanding that, so I
would withhold my judgment until I see what changes to the existing
code are needed to support that configuration.  I have no real
knowledge of how Clang supports native MS-Windows; I don't even know
what they use for the runtime library.  So I cannot envision up front
what kind of changes will be needed to produce a successful native
MS-Windows build of Emacs using Clang.

Message #29 received at 53377 <at> debbugs.gnu.org (full text, mbox):

From: Eli Zaretskii <eliz <at> gnu.org>
To: Óscar Fuentes <ofv <at> wanadoo.es>
Cc: luangruo <at> yahoo.com, 53377 <at> debbugs.gnu.org
Subject: Re: bug#53377: 28.0.91: Build fails on MSYS2/MinGW-w64 with Clang
Date: Thu, 20 Jan 2022 09:54:35 +0200

> From: Óscar Fuentes <ofv <at> wanadoo.es>
> Date: Thu, 20 Jan 2022 04:59:55 +0100
> Cc: 53377 <at> debbugs.gnu.org
> 
> Besides, Clang has its own debugger.

Last time I looked, the LLVM debugger didn't work well on Windows, and
was definitely not a good idea as a tool to debug Emacs, because it
didn't support the .gdbinit commands (which are essential for any
reasonable debugging of the Emacs C code), and didn't support GDB/MI
(which means you cannot use it with "M-x gdb" from Emacs).

Unless this has changed drastically since I looked, my advice to you
is to use GDB for debugging the Emacs C code, otherwise you will have
a whole additional level of difficulties in making this port happen.

Message #34 received at 53377 <at> debbugs.gnu.org (full text, mbox):

From: Richard Stallman <rms <at> gnu.org>
To: Eli Zaretskii <eliz <at> gnu.org>
Cc: luangruo <at> yahoo.com, ofv <at> wanadoo.es, 53377 <at> debbugs.gnu.org
Subject: Re: bug#53377: 28.0.91: Build fails on MSYS2/MinGW-w64 with Clang
Date: Thu, 20 Jan 2022 23:14:20 -0500

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

  > > Do we really want to support building Emacs with Clang on MS-Windows?

  > It depends on what would such support require.  Judging by what I see
  > in the original report, we are very far from understanding that, so I
  > would withhold my judgment until I see what changes to the existing
  > code are needed to support that configuration.

Even if it could be done, that doesn't imply we ought to put any
effort into it.

-- 
Dr Richard Stallman (https://stallman.org)
Chief GNUisance of the GNU Project (https://gnu.org)
Founder, Free Software Foundation (https://fsf.org)
Internet Hall-of-Famer (https://internethalloffame.org)

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.