clisp-devel — closed list for developers and testers

Re: New Cygwin maintainer

[Ken B. <kb...@co...>]

On 3/16/2015 2:28 PM, Blake McBride wrote:
> I noticed dynamic modules doesn't work on my 64 bit Linux either.  I too
> am using the respository head.  Any chances your patch will fix my
> problem too?
>
> If so, where do I get your patch?

The patches are pretty Cygwin-specific, but maybe there's an idea in 
there that could help you.  I'm attaching the current version.  (It 
could still change after further testing.)

Ken

Re: New Cygwin maintainer

[Blake M. <bl...@mc...>]

I noticed dynamic modules doesn't work on my 64 bit Linux either.  I too am
using the respository head.  Any chances your patch will fix my problem too?

If so, where do I get your patch?

Thanks!

Blake McBride


On Fri, Mar 13, 2015 at 3:24 PM, Ken Brown <kb...@co...> wrote:

> On 3/6/2015 11:58 AM, Ken Brown wrote:
> > I used the configure option
> >
> >     --without-dynamic-modules
> >
> > because dynamic modules currently do not work on Cygwin.  My next goal
> > is to try to fix this.  But I know that Reini worked on it without
> > success, so I'm not very optimistic.
>
> Just to follow up on this, it turned out that I was able to get it
> working after all.  In case there are Cygwin users on this list who
> would like to try it, I've just updated Cygwin's clisp package:
>
>    https://cygwin.com/ml/cygwin-announce/2015-03/msg00025.html
>
> I'll wait a couple weeks and then submit my patches if there are no
> problems reported.
>
> Ken
>
>
> ------------------------------------------------------------------------------
> Dive into the World of Parallel Programming The Go Parallel Website,
> sponsored
> by Intel and developed in partnership with Slashdot Media, is your hub for
> all
> things parallel software development, from weekly thought leadership blogs
> to
> news, videos, case studies, tutorials and more. Take a look and join the
> conversation now. http://goparallel.sourceforge.net/
> _______________________________________________
> clisp-devel mailing list
> cli...@li...
> https://lists.sourceforge.net/lists/listinfo/clisp-devel
>

Re: New Cygwin maintainer

[Ken B. <kb...@co...>]

On 3/6/2015 11:58 AM, Ken Brown wrote:
> I used the configure option
>
>     --without-dynamic-modules
>
> because dynamic modules currently do not work on Cygwin.  My next goal
> is to try to fix this.  But I know that Reini worked on it without
> success, so I'm not very optimistic.

Just to follow up on this, it turned out that I was able to get it 
working after all.  In case there are Cygwin users on this list who 
would like to try it, I've just updated Cygwin's clisp package:

   https://cygwin.com/ml/cygwin-announce/2015-03/msg00025.html

I'll wait a couple weeks and then submit my patches if there are no 
problems reported.

Ken

Re: New Cygwin maintainer

[Jerry J. <log...@gm...>]

On Fri, Mar 6, 2015 at 2:39 PM, Ken Brown <kb...@co...> wrote:
> On 3/6/2015 3:17 PM, Sam Steingold wrote:
>> Jerry, your patches have not been applied because, alas, CLISP has
>> fallen out of maintenance. We are sorry about that.

I understand.  I'm an XEmacs developer. :-)

>> Ken, Jerry,
>> If you are interested, we can make either of you (or both) the new
>> maintainers.
>
>
> I'm not qualified.  I'm not even a clisp user; I'm just working on build
> issues.  But I'd be glad to help in any way I can if someone steps forward
> to be the new maintainer.

I regret that I don't really have sufficient free time to take on
another open source project.  But I'm with Ken.  If somebody does step
forward to take up maintenance, I will help with Fedora build and
integration issues, and supply the occasional patch as needed.
-- 
Jerry James
http://www.jamezone.org/

Re: New Cygwin maintainer

[Ken B. <kb...@co...>]

On 3/6/2015 3:17 PM, Sam Steingold wrote:
>> * Ken Brown <xoebja@pbearyy.rqh> [2015-02-24 17:06:49 -0500]:
>>
>> I have just taken over from Reini Urban as clisp maintainer for
>> Cygwin.
>
> Thank you!
>
>> I can build the tip of the Mercurial repository on both 32-bit and
>> 64-bit Cygwin [*], but several patches are required (not all of which
>> are Cygwin-specific).  I would like to get these patches applied.  How
>> should I proceed?  Should I just send patches to this list?  Or do you
>> prefer bug reports with patches attached?
>
> Sorry about nor replying right away.
> Please submit your patches to
> https://sourceforge.net/p/clisp/patches/
> (you will need an SF account).

OK.

> Jerry, your patches have not been applied because, alas, CLISP has
> fallen out of maintenance. We are sorry about that.
>
> Ken, Jerry,
> If you are interested, we can make either of you (or both) the new
> maintainers.

I'm not qualified.  I'm not even a clisp user; I'm just working on build issues. 
  But I'd be glad to help in any way I can if someone steps forward to be the 
new maintainer.

Ken

Re: New Cygwin maintainer

[Sam S. <sd...@gn...>]

> * Ken Brown <xoebja@pbearyy.rqh> [2015-02-24 17:06:49 -0500]:
>
> I have just taken over from Reini Urban as clisp maintainer for
> Cygwin.

Thank you!

> I can build the tip of the Mercurial repository on both 32-bit and
> 64-bit Cygwin [*], but several patches are required (not all of which
> are Cygwin-specific).  I would like to get these patches applied.  How
> should I proceed?  Should I just send patches to this list?  Or do you
> prefer bug reports with patches attached?

Sorry about nor replying right away.
Please submit your patches to
https://sourceforge.net/p/clisp/patches/
(you will need an SF account).

Jerry, your patches have not been applied because, alas, CLISP has
fallen out of maintenance. We are sorry about that.

Ken, Jerry,
If you are interested, we can make either of you (or both) the new
maintainers.
(If either you or Ken is a student, you can apply to GSoC 2015 to do
that and get paid in the process. The success indicator will be
releasing CLISP 2.50 for Linux, Windows, Cygwin, MacOSX, and at least
one of FreeBSD, OpenBSD and NetBSD).

Thanks.

-- 
Sam Steingold (http://sds.podval.org/) on darwin Ns 10.3.1344
http://www.childpsy.net/ http://iris.org.il http://americancensorship.org
http://think-israel.org http://ffii.org http://honestreporting.com
My inferiority complex is not as good as yours.

Re: [PATCH 05/12] modules/libsvm/libsvm.lisp (svm_destroy_model): Fix prototype

[Jerry J. <log...@gm...>]

On Fri, Mar 6, 2015 at 10:00 AM, Ken Brown <kb...@co...> wrote:
> # HG changeset patch
> # User Ken Brown <kb...@co...>
> # Date 1425569108 18000
> #      Thu Mar 05 10:25:08 2015 -0500
> # Node ID c90aa4bb8c75be0942f2196dbbc477de56deede1
> # Parent  c89e1966ad491619b813bde15de5f1cb0418933a
> * modules/libsvm/libsvm.lisp (svm_destroy_model): Fix prototype
>
> diff -r c89e1966ad49 -r c90aa4bb8c75 modules/libsvm/libsvm.lisp
> --- a/modules/libsvm/libsvm.lisp        Thu Mar 05 10:23:05 2015 -0500
> +++ b/modules/libsvm/libsvm.lisp        Thu Mar 05 10:25:08 2015 -0500
> @@ -31,7 +31,7 @@
>  (c-lines "
>  #if !defined(HAVE_SVM_DESTROY_MODEL)
>  # if defined(HAVE_SVM_FREE_AND_DESTROY_MODEL)
> -void svm_destroy_model (svm_model *model){ svm_free_and_destroy_model(&model); }
> +void svm_destroy_model (struct svm_model *model){ svm_free_and_destroy_model(&model); }
>  # else
>  #  error No svm_destroy_model and no svm_free_and_destroy_model!
>  # endif

This one was submitted to the patch tracker 27 Sep 2012:
https://sourceforge.net/p/clisp/patches/34/.  I've got a few other
patches sitting there, untouched for years. :-(

Regards,
-- 
Jerry James
http://www.jamezone.org/

[PATCH 12/12] Work around gcc problem on 64-bit Cygwin; *FOR LOCAL USE ONLY*

[Ken B. <kb...@co...>]

# HG changeset patch
# User Ken Brown <kb...@co...>
# Date 1425654503 18000
#      Fri Mar 06 10:08:23 2015 -0500
# Node ID ad1b90a12676777818a170ede793f34765dcffe6
# Parent  3ee52afa4969411b4b7d61b2232c9274bcd1bce4
* src/makemake.in (BASE_MODULES): Work around gcc problem on 64-bit Cygwin

diff -r 3ee52afa4969 -r ad1b90a12676 src/makemake.in
--- a/src/makemake.in   Fri Mar 06 09:41:46 2015 -0500
+++ b/src/makemake.in   Fri Mar 06 10:08:23 2015 -0500
@@ -658,6 +658,11 @@
      ;;
 esac

+# Temporary workaround for gcc problem on 64-bit Cygwin
+if [ "$HSYSOS" = cygwin ] && [ "$HSYS" = x86_64 ]; then
+  BASE_MODULES="i18n syscalls"
+fi
+
 if [ "${with_gmalloc}" = yes ]; then GMALLOC=gmalloc; else GMALLOC=""; fi

 if [ $CROSS = true ] ; then

[PATCH 11/12] modules/syscalls/configure: Update; *maybe configure should be regenerated instead*

[Ken B. <kb...@co...>]

# HG changeset patch
# User Ken Brown <kb...@co...>
# Date 1425652906 18000
#      Fri Mar 06 09:41:46 2015 -0500
# Node ID 3ee52afa4969411b4b7d61b2232c9274bcd1bce4
# Parent  45b9d318b7c64782ca1d7cc88ac7ba50a9c97e30
* modules/syscalls/configure: Update

diff -r 45b9d318b7c6 -r 3ee52afa4969 modules/syscalls/configure
--- a/modules/syscalls/configure        Fri Mar 06 09:33:48 2015 -0500
+++ b/modules/syscalls/configure        Fri Mar 06 09:41:46 2015 -0500
@@ -4798,7 +4798,12 @@
 done

 if test "$ac_cv_header_shlobj_h" = yes ; then
-   LIBS=${LIBS}' -luser32 -lole32 -loleaut32 -luuid -lversion';
+   case "$ac_cv_build" in
+   *-cygwin)
+     LIBS=${LIBS}' -luser32 -lole32 -loleaut32 -L/usr/lib/w32api -luuid -lversion';;
+   *)
+     LIBS=${LIBS}' -luser32 -lole32 -loleaut32 -luuid -lversion';;
+   esac
 fi
 # Check whether --enable-largefile was given.
 if test "${enable_largefile+set}" = set; then :

[PATCH 10/12] src/configure: Update; *maybe configure should be regenerated instead*

[Ken B. <kb...@co...>]

# HG changeset patch
# User Ken Brown <kb...@co...>
# Date 1425652428 18000
#      Fri Mar 06 09:33:48 2015 -0500
# Node ID 45b9d318b7c64782ca1d7cc88ac7ba50a9c97e30
# Parent  92c3ad94be9ada0750bd1ea96afc8101e0917af6
* src/configure: Update

diff -r 92c3ad94be9a -r 45b9d318b7c6 src/configure
--- a/src/configure     Thu Mar 05 12:30:28 2015 -0500
+++ b/src/configure     Fri Mar 06 09:33:48 2015 -0500
@@ -34298,11 +34298,6 @@
 #include <unistd.h>
 #endif
 #include <signal.h>
-#ifdef __CYGWIN32__
-/* On Cygwin32 version 18, this test would hang (SIGALRM not being signalled).
- * Let it fail instead. */
-#error "better fail than hang"
-#endif
 volatile int gotsig=0;
 void sigalrm_handler() { gotsig=1; }
 int got_sig () { return gotsig; }
@@ -34352,11 +34347,6 @@
 #include <unistd.h>
 #endif
 #include <signal.h>
-#ifdef __CYGWIN32__
-/* On Cygwin32 version 18, this test would hang (SIGALRM not being signalled).
- * Let it fail instead. */
-#error "better fail than hang"
-#endif
 volatile int gotsig=0;
 volatile int wasblocked=0;
 typedef void (*signal_handler_t) (int);
@@ -34417,11 +34407,6 @@
 #include <unistd.h>
 #endif
 #include <signal.h>
-#ifdef __CYGWIN32__
-/* On Cygwin32 version 18, this test would hang (SIGALRM not being signalled).
- * Let it fail instead. */
-#error "better fail than hang"
-#endif
 volatile int gotsig=0;
 volatile int somewereblocked=0;
 typedef void (*signal_handler_t) (int);
@@ -34497,11 +34482,6 @@
 #include <unistd.h>
 #endif
 #include <signal.h>
-#ifdef __CYGWIN32__
-/* On Cygwin32 version 18, this test would hang (SIGALRM not being signalled).
- * Let it fail instead. */
-#error "better fail than hang"
-#endif
 typedef void (*signal_handler_t) (int);
 signal_handler_t mysignal (int sig, signal_handler_t handler)
 { struct sigaction old_sa;
@@ -34561,11 +34541,6 @@
 #include <unistd.h>
 #endif
 #include <signal.h>
-#ifdef __CYGWIN32__
-/* On Cygwin32 version 18, this test would hang (SIGALRM not being signalled).
- * Let it fail instead. */
-#error "better fail than hang"
-#endif
 typedef void (*signal_handler_t) (int);
 signal_handler_t mysignal (int sig, signal_handler_t handler)
 { struct sigaction old_sa;

[PATCH 09/12]

[Ken B. <kb...@co...>]

# HG changeset patch
# User Ken Brown <kb...@co...>
# Date 1425576628 18000
#      Thu Mar 05 12:30:28 2015 -0500
# Node ID 74b47369b904fba3b89b479173268d5382ab7189
# Parent  2130cb689a5d63ae7e97facbc2336c6517f64247
* src/lispbibl.d (ULONGLONG): Avoid conflict with w32 typedef on 64-bit Cygwin

diff -r 2130cb689a5d -r 74b47369b904 src/lispbibl.d
--- a/src/lispbibl.d    Thu Mar 05 11:35:30 2015 -0500
+++ b/src/lispbibl.d    Thu Mar 05 12:30:28 2015 -0500
@@ -1096,7 +1096,7 @@
 #else
   #error No 32 bit integer type? -- Which Integer-type has 32 Bit?
 #endif
-#if (long_bitsize==64)
+#if (long_bitsize==64) && !defined(UNIX_CYGWIN32)
   typedef long           SLONGLONG;
   typedef unsigned long  ULONGLONG;
   #ifndef HAVE_LONG_LONG_INT
@@ -1132,7 +1132,7 @@
 %%   emit_typedef("int","SLONG");
 %%   emit_typedef("unsigned int","ULONG");
 %% #endif
-%% #if (long_bitsize==64)
+%% #if (long_bitsize==64) && !defined(UNIX_CYGWIN32)
 %%   emit_typedef("long","SLONGLONG");
 %%   emit_typedef("unsigned long","ULONGLONG");
 %% #elif defined(MICROSOFT)

[PATCH 08/12] Use __CYGWIN__ instead of __CYGWIN32__ to detect Cygwin

[Ken B. <kb...@co...>]

# HG changeset patch
# User Ken Brown <kb...@co...>
# Date 1425573330 18000
#      Thu Mar 05 11:35:30 2015 -0500
# Node ID 2130cb689a5d63ae7e97facbc2336c6517f64247
# Parent  ce68b45a1070376b75abb61d8be7ee908e949805
Use __CYGWIN__ instead of __CYGWIN32__ to detect Cygwin

__CYGWIN32__ is deprecated and, more importantly, is not defined on
64-bit Cygwin.

diff -r ce68b45a1070 -r 2130cb689a5d src/ari80386.d
--- a/src/ari80386.d    Thu Mar 05 11:29:12 2015 -0500
+++ b/src/ari80386.d    Thu Mar 05 11:35:30 2015 -0500
@@ -52,7 +52,7 @@
     #endif
     # Alignment. Note that some assemblers need ".align 3,0x90" whereas other
     # assemblers don't like this syntax. So we put in the "nop"s by hand.
-    #if defined(ASM_UNDERSCORE) && !(defined(__CYGWIN32__) || defined(__MINGW32__))
+    #if defined(ASM_UNDERSCORE) && !(defined(__CYGWIN__) || defined(__MINGW32__))
       # BSD syntax assembler
       #define ALIGN  .align 3
     #else
diff -r ce68b45a1070 -r 2130cb689a5d src/ari80386.msvc.c
--- a/src/ari80386.msvc.c       Thu Mar 05 11:29:12 2015 -0500
+++ b/src/ari80386.msvc.c       Thu Mar 05 11:35:30 2015 -0500
@@ -52,7 +52,7 @@
     #endif


-    #if defined(ASM_UNDERSCORE) && !(defined(__CYGWIN32__) || defined(__MINGW32__))
+    #if defined(ASM_UNDERSCORE) && !(defined(__CYGWIN__) || defined(__MINGW32__))

       #define ALIGN  .align 3
     #else
diff -r ce68b45a1070 -r 2130cb689a5d src/asmi386.h
--- a/src/asmi386.h     Thu Mar 05 11:29:12 2015 -0500
+++ b/src/asmi386.h     Thu Mar 05 11:35:30 2015 -0500
@@ -174,10 +174,10 @@
 #define REPZ repz ;
 #endif
 #define REP rep ;
-#if defined(BSD_SYNTAX) && !(defined(__CYGWIN32__) || defined(__MINGW32__))
+#if defined(BSD_SYNTAX) && !(defined(__CYGWIN__) || defined(__MINGW32__))
 #define ALIGN(log) .align log,0x90
 #endif
-#if defined(ELF_SYNTAX) || defined(__CYGWIN32__) || defined(__MINGW32__)
+#if defined(ELF_SYNTAX) || defined(__CYGWIN__) || defined(__MINGW32__)
 #define ALIGN(log) .align 1<<log
 #endif
 #endif
diff -r ce68b45a1070 -r 2130cb689a5d src/asmi386.hh
--- a/src/asmi386.hh    Thu Mar 05 11:29:12 2015 -0500
+++ b/src/asmi386.hh    Thu Mar 05 11:35:30 2015 -0500
@@ -174,10 +174,10 @@
 #define REPZ repz ;
 #endif
 #define REP rep ;
-#if defined(BSD_SYNTAX) && !(defined(__CYGWIN32__) || defined(__MINGW32__))
+#if defined(BSD_SYNTAX) && !(defined(__CYGWIN__) || defined(__MINGW32__))
 #define ALIGN(log) .align log,0x90
 #endif
-#if defined(ELF_SYNTAX) || defined(__CYGWIN32__) || defined(__MINGW32__)
+#if defined(ELF_SYNTAX) || defined(__CYGWIN__) || defined(__MINGW32__)
 #define ALIGN(log) .align 1<<log
 #endif
 #endif
diff -r ce68b45a1070 -r 2130cb689a5d src/lispbibl.d
--- a/src/lispbibl.d    Thu Mar 05 11:29:12 2015 -0500
+++ b/src/lispbibl.d    Thu Mar 05 11:35:30 2015 -0500
@@ -84,11 +84,11 @@
  IBM-PC/386   any                NetBSD 0.9 (UNIX BSD 4.3)     GNU           unix, __NetBSD__, i386, __GNUC__
  IBM-PC/386   any                FreeBSD 4.0 (UNIX BSD 4.4)    GNU           unix, __FreeBSD__, i386, __GNUC__
  IBM-PC/386   any                EMX 0.9c (UNIXlike on OS/2)   GNU           [unix,] i386, __GNUC__, __EMX__
- IBM-PC/386   any                Cygwin32 on WinNT/Win95       GNU           _WIN32, __WINNT__, __CYGWIN32__, __POSIX__, _X86_, i386, __GNUC__
+ IBM-PC/386   any                Cygwin32 on WinNT/Win95       GNU           _WIN32, __WINNT__, __CYGWIN__, __POSIX__, _X86_, i386, __GNUC__
  IBM-PC/386   any                Mingw32 on WinNT/Win95        GNU           _WIN32, __WINNT__, __MINGW32__, _X86_, i386, __GNUC__
  IBM-PC/386   any                WinNT/Win95                   MSVC          _WIN32, _M_IX86, _MSC_VER
  IBM-PC/386   any                WinNT/Win95                   Borland 5.0   __WIN32__, _M_IX86, __TURBOC__, __BORLANDC__
- IBM-PC/386   any                WinNT/Win95 and Cygwin32      GNU           _WIN32, __WINNT__, __CYGWIN32__, __POSIX__, __i386__, _X86_, __GNUC__
+ IBM-PC/386   any                WinNT/Win95 and Cygwin32      GNU           _WIN32, __WINNT__, __CYGWIN__, __POSIX__, __i386__, _X86_, __GNUC__
  IBM-PC/586   any                BeOS 5                        GNU           __BEOS__, __INTEL__, __i386__, _X86_, __GNUC__
  IBM-PC/586   any                HP NUE/ski, Linux             GNU           unix, linux, __ia64[__], __GNUC__, __LP64__
  RM400        Siemens-Nixdorf    SINIX-N 5.42                  c89           unix, mips, MIPSEB, host_mips, sinix, SNI, _XPG_IV
@@ -121,7 +121,7 @@
   #define PC386 /* IBMPC-compatible with 80386/80486-processor */
 #endif
 #ifdef GENERIC_UNIX
-  #if (defined(unix) && (defined(linux) || defined(__CYGWIN32__) || defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__) || defined(__DragonFly__)) && (defined(i386) || defined(__i386__) || defined(__x86_64__) || defined(__amd64__)))
+  #if (defined(unix) && (defined(linux) || defined(__CYGWIN__) || defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__) || defined(__DragonFly__)) && (defined(i386) || defined(__i386__) || defined(__x86_64__) || defined(__amd64__)))
     #define PC386
   #endif
   #if (defined(sun) && defined(unix) && defined(sparc))
@@ -295,7 +295,7 @@
   #ifdef AMIX
     #define UNIX_AMIX  /* Amiga UNIX */
   #endif
-  #ifdef __CYGWIN32__
+  #ifdef __CYGWIN__
     #define UNIX_CYGWIN32  /* Cygwin32 (UNIXlike on WinNT/Win95) */
   #endif
   #ifdef __BEOS__
diff -r ce68b45a1070 -r 2130cb689a5d src/sp80386.d
--- a/src/sp80386.d     Thu Mar 05 11:29:12 2015 -0500
+++ b/src/sp80386.d     Thu Mar 05 11:35:30 2015 -0500
@@ -12,7 +12,7 @@
     #endif
     # Alignment. Note that some assemblers need ".align 3,0x90" whereas other
     # assemblers don't like this syntax. So we put in the "nop"s by hand.
-    #if defined(ASM_UNDERSCORE) && !(defined(__CYGWIN32__) || defined(__MINGW32__))
+    #if defined(ASM_UNDERSCORE) && !(defined(__CYGWIN__) || defined(__MINGW32__))
       # BSD syntax assembler
       #define ALIGN  .align 3
     #else
diff -r ce68b45a1070 -r 2130cb689a5d src/sp80386.msvc.c
--- a/src/sp80386.msvc.c        Thu Mar 05 11:29:12 2015 -0500
+++ b/src/sp80386.msvc.c        Thu Mar 05 11:35:30 2015 -0500
@@ -12,7 +12,7 @@
     #endif


-    #if defined(ASM_UNDERSCORE) && !(defined(__CYGWIN32__) || defined(__MINGW32__))
+    #if defined(ASM_UNDERSCORE) && !(defined(__CYGWIN__) || defined(__MINGW32__))

       #define ALIGN  .align 3
     #else

[PATCH 07/12] src/m4/signal.m4: Remove obsolete Cygwin-specific code

[Ken B. <kb...@co...>]

# HG changeset patch
# User Ken Brown <kb...@co...>
# Date 1425572886 18000
#      Thu Mar 05 11:28:06 2015 -0500
# Node ID 3156692cdd6e5f8d535c730dd499ddf750c90621
# Parent  3e0a47cc519e131f658131387a5dd118cacff85d
* src/m4/signal.m4: Remove obsolete Cygwin-specific code

diff -r 3e0a47cc519e -r 3156692cdd6e src/m4/signal.m4
--- a/src/m4/signal.m4  Thu Mar 05 10:48:49 2015 -0500
+++ b/src/m4/signal.m4  Thu Mar 05 11:28:06 2015 -0500
@@ -20,11 +20,6 @@
 #include <unistd.h>
 #endif
 #include <signal.h>
-#ifdef __CYGWIN32__
-/* On Cygwin32 version 18, this test would hang (SIGALRM not being signalled).
- * Let it fail instead. */
-#error "better fail than hang"
-#endif
 volatile int gotsig=0;
 void sigalrm_handler() { gotsig=1; }
 int got_sig () { return gotsig; }
@@ -55,11 +50,6 @@
 #include <unistd.h>
 #endif
 #include <signal.h>
-#ifdef __CYGWIN32__
-/* On Cygwin32 version 18, this test would hang (SIGALRM not being signalled).
- * Let it fail instead. */
-#error "better fail than hang"
-#endif
 volatile int gotsig=0;
 volatile int wasblocked=0;
 typedef void (*signal_handler_t) (int);
@@ -101,11 +91,6 @@
 #include <unistd.h>
 #endif
 #include <signal.h>
-#ifdef __CYGWIN32__
-/* On Cygwin32 version 18, this test would hang (SIGALRM not being signalled).
- * Let it fail instead. */
-#error "better fail than hang"
-#endif
 volatile int gotsig=0;
 volatile int somewereblocked=0;
 typedef void (*signal_handler_t) (int);
@@ -156,11 +141,6 @@
 #include <unistd.h>
 #endif
 #include <signal.h>
-#ifdef __CYGWIN32__
-/* On Cygwin32 version 18, this test would hang (SIGALRM not being signalled).
- * Let it fail instead. */
-#error "better fail than hang"
-#endif
 typedef void (*signal_handler_t) (int);
 signal_handler_t mysignal (int sig, signal_handler_t handler)
 { struct sigaction old_sa;
@@ -202,11 +182,6 @@
 #include <unistd.h>
 #endif
 #include <signal.h>
-#ifdef __CYGWIN32__
-/* On Cygwin32 version 18, this test would hang (SIGALRM not being signalled).
- * Let it fail instead. */
-#error "better fail than hang"
-#endif
 typedef void (*signal_handler_t) (int);
 signal_handler_t mysignal (int sig, signal_handler_t handler)
 { struct sigaction old_sa;

[PATCH 06/12] modules/gtk2/gtk-server.cfg: Add Cygwin libs (commented out)

[Ken B. <kb...@co...>]

# HG changeset patch
# User Ken Brown <kb...@co...>
# Date 1425570529 18000
#      Thu Mar 05 10:48:49 2015 -0500
# Node ID 3e0a47cc519e131f658131387a5dd118cacff85d
# Parent  c90aa4bb8c75be0942f2196dbbc477de56deede1
* modules/gtk2/gtk-server.cfg: Add Cygwin libs (commented out)

diff -r c90aa4bb8c75 -r 3e0a47cc519e modules/gtk2/gtk-server.cfg
--- a/modules/gtk2/gtk-server.cfg       Thu Mar 05 10:25:08 2015 -0500
+++ b/modules/gtk2/gtk-server.cfg       Thu Mar 05 10:48:49 2015 -0500
@@ -64,6 +64,19 @@
 #
 #---------------------------------------------------------------------------
 #
+# Cygwin
+#
+# GTK2 backend:
+# GTK_LIB_NAME = cyggtk-x11-2.0-0.dll
+# GDK_LIB_NAME = cyggdk-x11-2.0-0.dll
+# GLIB_LIB_NAME = cygglib-2.0-0.dll
+# GOBJECT_LIB_NAME = cyggobject-2.0-0.dll
+# ATK_LIB_NAME = cygatk-1.0-0.dll
+# PANGO_LIB_NAME = cygpango-1.0-0.dll
+# PIXBUF_LIB_NAME = cyggdk_pixbuf_xlib-2.0-0.dll
+#
+#---------------------------------------------------------------------------
+#
 # Directory where to put the logfile if logging is enabled.
 # In Windows, also use a slash forward (/) to separate directory's!
 #

[PATCH 05/12] modules/libsvm/libsvm.lisp (svm_destroy_model): Fix prototype

[Ken B. <kb...@co...>]

# HG changeset patch
# User Ken Brown <kb...@co...>
# Date 1425569108 18000
#      Thu Mar 05 10:25:08 2015 -0500
# Node ID c90aa4bb8c75be0942f2196dbbc477de56deede1
# Parent  c89e1966ad491619b813bde15de5f1cb0418933a
* modules/libsvm/libsvm.lisp (svm_destroy_model): Fix prototype

diff -r c89e1966ad49 -r c90aa4bb8c75 modules/libsvm/libsvm.lisp
--- a/modules/libsvm/libsvm.lisp        Thu Mar 05 10:23:05 2015 -0500
+++ b/modules/libsvm/libsvm.lisp        Thu Mar 05 10:25:08 2015 -0500
@@ -31,7 +31,7 @@
 (c-lines "
 #if !defined(HAVE_SVM_DESTROY_MODEL)
 # if defined(HAVE_SVM_FREE_AND_DESTROY_MODEL)
-void svm_destroy_model (svm_model *model){ svm_free_and_destroy_model(&model); }
+void svm_destroy_model (struct svm_model *model){ svm_free_and_destroy_model(&model); }
 # else
 #  error No svm_destroy_model and no svm_free_and_destroy_model!
 # endif

[PATCH 04/12] Ensure that `-luuid' finds the w32 library

[Ken B. <kb...@co...>]

# HG changeset patch
# User Ken Brown <kb...@co...>
# Date 1425566518 18000
#      Thu Mar 05 09:41:58 2015 -0500
# Node ID 27a4d1cc9090b02c8f388714b83fd8184cad6ca7
# Parent  df85b49c7315d4c365f32455f4fdeee2893c48b4
* modules/syscalls/configure.in [cygwin]: Ensure that `-luuid' finds the w32 library

diff -r df85b49c7315 -r 27a4d1cc9090 modules/syscalls/configure.in
--- a/modules/syscalls/configure.in     Wed Mar 04 22:05:35 2015 -0500
+++ b/modules/syscalls/configure.in     Thu Mar 05 09:41:58 2015 -0500
@@ -18,7 +18,12 @@
 pwd.h sys/unistd.h syslog.h signal.h grp.h crypt.h utmpx.h shlobj.h dnl
 sys/statvfs.h sys/statfs.h sys/param.h ftw.h)
 if test "$ac_cv_header_shlobj_h" = yes ; then
-   LIBS=${LIBS}' -luser32 -lole32 -loleaut32 -luuid -lversion';
+   case "$ac_cv_build" in
+   *-cygwin)
+     LIBS=${LIBS}' -luser32 -lole32 -loleaut32 -L/usr/lib/w32api -luuid -lversion';;
+   *)
+     LIBS=${LIBS}' -luser32 -lole32 -loleaut32 -luuid -lversion';;
+   esac
 fi
 AC_SYS_LARGEFILE    dnl ensure 64 bit size_t for AC_CHECK_SIZEOF below
 AC_CHECK_SIZEOF(off_t)

[PATCH 03/12] Use LINUX_NOEXEC_HEAPCODES on 32-bit Cygwin

[Ken B. <kb...@co...>]

# HG changeset patch
# User Ken Brown <kb...@co...>
# Date 1425524735 18000
#      Wed Mar 04 22:05:35 2015 -0500
# Node ID df85b49c7315d4c365f32455f4fdeee2893c48b4
# Parent  34be157e71d92a16c20bb601828cee54156ad2a5
* src/lispbibl.d: Use LINUX_NOEXEC_HEAPCODES on 32-bit Cygwin

The heap on Cygwin is now in high memory, so STANDARD_HEAPCODES will
no longer work.

diff -r 34be157e71d9 -r df85b49c7315 src/lispbibl.d
--- a/src/lispbibl.d    Wed Mar 04 21:54:40 2015 -0500
+++ b/src/lispbibl.d    Wed Mar 04 22:05:35 2015 -0500
@@ -2506,7 +2506,7 @@
      malloc results (and hence also of shared libraries) are randomized;
      only the code address is fixed around 0x1C000000 and the stack address
      is around 0xCF000000. In this case, we also use LINUX_NOEXEC_HEAPCODES. */
-    #if (defined(I80386) && defined(UNIX_LINUX)) || (defined(I80386) && defined(UNIX_OPENBSD) && defined(ADDRESS_RANGE_RANDOMIZED))
+    #if (defined(I80386) && defined(UNIX_LINUX)) || (defined(I80386) && defined(UNIX_OPENBSD) && defined(ADDRESS_RANGE_RANDOMIZED)) || (defined(I80386) && defined(UNIX_CYGWIN32))
       #define LINUX_NOEXEC_HEAPCODES
     #else
       #define STANDARD_HEAPCODES

[PATCH 02/12] Restore declaration of ioctl

[Ken B. <kb...@co...>]

# HG changeset patch
# User Ken Brown <kb...@co...>
# Date 1425524080 18000
#      Wed Mar 04 21:54:40 2015 -0500
# Node ID 34be157e71d92a16c20bb601828cee54156ad2a5
# Parent  9a43f7dac04375804add3f16955308f0a44cc9b8
Restore declaration of ioctl

* src/unix.d [HAVE_IOCTL]: Restore declaration of ioctl, which was
removed as part of revision 15470, leading to an
implicit-function-declaration warning in src/unixaux.d.

diff -r 9a43f7dac043 -r 34be157e71d9 src/unix.d
--- a/src/unix.d        Wed Mar 04 18:11:34 2015 -0500
+++ b/src/unix.d        Wed Mar 04 21:54:40 2015 -0500
@@ -433,6 +433,18 @@
 #elif defined(NEED_SYS_IOCTL_H)
   #include <sys/ioctl.h>
 #endif
+
+#ifdef HAVE_IOCTL
+  #ifdef IOCTL_DOTS
+    extern_C int ioctl (int fd, IOCTL_REQUEST_T request, ...); /* IOCTL(2) */
+    #define IOCTL_ARGUMENT_T  CADDR_T
+  #else
+    extern_C int ioctl (int fd, IOCTL_REQUEST_T request, IOCTL_ARGUMENT_T arg); /* IOCTL(2) */
+    /* 3rd argument is always cast to type IOCTL_ARGUMENT_T (usually CADDR_T): */
+    #define ioctl(fd,request,arg)  (ioctl)(fd,request,(IOCTL_ARGUMENT_T)(arg))
+  #endif
+#endif
+
 /* START_NO_BLOCK() & END_NO_BLOCK() should appear in pairs
    inside { NO_BLOCK_DECL(); ... };
  NO_BLOCK_DECL() should be before the first statement,

[PATCH 01/12] Work around namespace pollution in w32 headers

[Ken B. <kb...@co...>]

# HG changeset patch
# User Ken Brown <kb...@co...>
# Date 1425510694 18000
#      Wed Mar 04 18:11:34 2015 -0500
# Node ID 9a43f7dac04375804add3f16955308f0a44cc9b8
# Parent  5c63938ef493e06b089fd6868293f39496413b78
* src/unix.d [UNIX_CYGWIN32]: Work around namespace pollution in w32 headers

"Handle" is used in a w32 prototype, interfering with our definition
of it as a macro.

diff -r 5c63938ef493 -r 9a43f7dac043 src/unix.d
--- a/src/unix.d        Wed Oct 22 12:04:12 2014 -0400
+++ b/src/unix.d        Wed Mar 04 18:11:34 2015 -0500
@@ -623,7 +623,10 @@
 /* Interpretation of FILETIME structure: */
 #ifdef UNIX_CYGWIN32
   #define WIN32_LEAN_AND_MEAN
+  #pragma push_macro ("Handle")
+  #undef Handle
   #include <windows.h>
+  #pragma pop_macro ("Handle")
   #undef WIN32
   extern long time_t_from_filetime (const FILETIME * ptr);
   extern void time_t_to_filetime (time_t time_in, FILETIME * out);

Re: New Cygwin maintainer

[Ken B. <kb...@co...>]

On 2/24/2015 5:06 PM, Ken Brown wrote:
> I have just taken over from Reini Urban as clisp maintainer for Cygwin.
>    I can build the tip of the Mercurial repository on both 32-bit and
> 64-bit Cygwin [*], but several patches are required (not all of which
> are Cygwin-specific).  I would like to get these patches applied.  How
> should I proceed?  Should I just send patches to this list?  Or do you
> prefer bug reports with patches attached?

Since no one answered this, and it's easier to send patches than to make 
bug reports, I'll send a sequence of patches in separate emails.

The are 12 patches.  The first 9 are intended to be applied to the 
Mercurial repository.  Patches 10 and 11 just update two configure 
scripts so that the tip still builds.  Presumably the maintainers should 
regenerate the configure scripts instead of applying the patches 
directly, but maybe it's convenient for them to apply the patches as a 
temporary measure.

Patch 12 is definitely *not* to be applied; it simply disables the 
building of regexp on 64-bit Cygwin because of the gcc problem mentioned 
below.  I'm sending it so that anyone who wants to try to replicate the 
build can apply it locally.

Here are some further remarks for people who want to replicate the build:

1. I used the configure option

   --without-dynamic-modules

because dynamic modules currently do not work on Cygwin.  My next goal 
is to try to fix this.  But I know that Reini worked on it without 
success, so I'm not very optimistic.

2. In the 64-bit case, I also used the following configure options:

   --ignore-absence-of-libsigsegv --without-ffcall

3. The build fails with mysterious error messages (on both 32-bit and 
64-bit Cygwin) unless the executables are built with a big enough stack. 
  I built with

   LDFLAGS="-Wl,--stack,0x800000"

4. In the 64-bit case, the only modules I could build were rawsock and 
dirkey.  All others that I tried failed to build either because they 
required ffcall or because they led to the gcc problem below.

Feedback on my patches is welcome.  If I need to sign a copyright 
assignment in order to get them applied, I'll be glad to do so.

Ken

> I ran into a gcc problem when trying to build several of the modules
> (including regexp):
>
>       https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64939

Re: New Cygwin maintainer => 64bit

[Ken B. <kb...@co...>]

On 3/2/2015 9:37 AM, Ken Brown wrote:
>> generational GC requires libsigsegv.
>
> Thanks.  The build on 32-bit Cygwin uses libsigsegv but not generational
> GC, so there's no difference here between the 32-bit and 64-bit
> situations.  I should probably look into why generational GC isn't used
> in the 32-bit case.

The problem turns out to be that clisp's configure doesn't like Cygwin's 
mmap.  I tried to build with -DHAVE_MMAP_ANON to see if this was just a 
problem with the configure test, but the build failed with a complaint 
about EXTRA-FILE-TYPES not being defined.  I didn't pursue it further.

Ken

Re: New Cygwin maintainer => 64bit

[Ken B. <kb...@co...>]

On 3/1/2015 8:24 PM, Sam Steingold wrote:
>> * Ken Brown <xoebja@pbearyy.rqh> [2015-02-25 17:32:40 -0500]:
>>
>>>    > 1. ffcall has not been ported to 64-bit Cygwin.
>>>    > 2. libsigsegv has not been ported to 64-bit Cygwin.
>>> Doesn't this affect GC ?
>>
>> I'm a clisp novice; I don't really know the connection between
>> libsigsegv and GC.
>
> generational GC requires libsigsegv.

Thanks.  The build on 32-bit Cygwin uses libsigsegv but not generational 
GC, so there's no difference here between the 32-bit and 64-bit 
situations.  I should probably look into why generational GC isn't used 
in the 32-bit case.

Ken

Re: New Cygwin maintainer => 64bit

[Sam S. <sd...@gn...>]

> * Ken Brown <xoebja@pbearyy.rqh> [2015-02-25 17:32:40 -0500]:
>
>>   > 1. ffcall has not been ported to 64-bit Cygwin.
>>   > 2. libsigsegv has not been ported to 64-bit Cygwin.
>> Doesn't this affect GC ?
>
> I'm a clisp novice; I don't really know the connection between 
> libsigsegv and GC.

generational GC requires libsigsegv.

-- 
Sam Steingold (http://sds.podval.org/) on darwin Ns 10.3.1344
http://www.childpsy.net/ http://palestinefacts.org http://www.dhimmitude.org
http://americancensorship.org http://camera.org http://www.memritv.org
Business is a game. If you know the rules, you can win, unless you follow them.

Re: New Cygwin maintainer => 64bit

[Ken B. <kb...@co...>]

On 2/25/2015 3:58 PM, Don Cohen wrote:
> Ken Brown writes:
>   > I have just taken over from Reini Urban as clisp maintainer for Cygwin.
>   >   I can build the tip of the Mercurial repository on both 32-bit and
>   > 64-bit Cygwin [*]
>
> Where can we get this to try out?

My build of clisp-2.49 is already in the Cygwin distribution.  If you 
want to try my build of the tip specifically, I can make it available.

>   > 1. ffcall has not been ported to 64-bit Cygwin.
>   > 2. libsigsegv has not been ported to 64-bit Cygwin.
> Doesn't this affect GC ?

I'm a clisp novice; I don't really know the connection between 
libsigsegv and GC.

> Or are you using a "memory model" that doesn't require it?

I'm just using whatever memory model is built by default.  It seems to 
work, so I haven't had to look into it in detail.

Ken

New Cygwin maintainer => 64bit

[<don...@is...>]

Ken Brown writes:
 > I have just taken over from Reini Urban as clisp maintainer for Cygwin. 
 >   I can build the tip of the Mercurial repository on both 32-bit and 
 > 64-bit Cygwin [*]

Where can we get this to try out?

 > 1. ffcall has not been ported to 64-bit Cygwin.
 > 2. libsigsegv has not been ported to 64-bit Cygwin.
Doesn't this affect GC ?
Or are you using a "memory model" that doesn't require it?