Re: [Axiom-developer] openpty patch again

[Gabriel Dos Reis]

address@hidden writes:

| I'm resubmitting a shorter version of a 300K patch I sent this
| morning.

I did not receive that patch...

| This time I generated the patch without the diffs for configure
| and a couple of Makefile.in files. I guess my version of configure is
| different from Gaby's, it basically changed every line in configure.

I'm using Autoconf 2.59; what is yours?

| I think I got a working build of axiom on Mac OS X 10.4 again.

Yay!

| http://www.hpcf.upr.edu/~humberto/axiom-macos-ppc-20061130.tar.gz
| 
| untar this somewhere, then point AXIOM at the
| target/powerpc-apple-darwin8.8.0 directory:
| 
| $ tar zxf axiom-macos-ppc-20061130.tar.gz
| $ export AXIOM=`pwd`/target/powerpc-apple-darwin8.8.0
| $ PATH=$AXIOM/bin:$PATH
| $ axiom
| 
| Success!
| 
| This was built from the svk mirror:
| 
| $ svk info
| Checkout Path: /Users/humberto/src/axiom/build-improvements
| Depot Path: /mirror/axiom/branches/build-improvements
| Revision: 328
| Last Changed Rev.: 327
| Mirrored From: https://svn.sourceforge.net/svnroot/axiom, Rev. 327
| Copied From: /axiom/trunk/axiom, Rev. 32
| Merged From: /axiom/trunk/axiom, Rev. 32
| 
| that I checked out on Tuesday. That revision builds on Mac OS X, but
| axiom won't work because sman and clef build without working pty support
| (AXIOMsys does work).
| 
| I reworked Waldek's better pty patch to work with the new
| config/axiom-c-macro.h file, and to check for the proper include file
| for openpty, at least on Mac and linux. I don't know if ptys work on MS
| Windows.

I don't think they, but I may be wrong -- I'm not a Windows developer
and I approach this whole stuff with a very engrained unix mind.

In my locat tree, I've divided the C runtime support (libspad.a) into 
five components:

   \begin{itemize}
   \item Core runtime support,
   \item interface with the operating system (filesystem, etc.),
   \item communication through sockets,
   \item user-interface through terminals, and
   \item graphics components.
   \end{itemize}

Core runtime support is hash.c and halloc.
interface with OS is anything havind to do with files, directories and
the like.
sockets is sockio-c.c
terminal-based ui is clef and sman
graphics is hyperdoc and, wel, graphics.

In my local tree it is possible to enable each individual component,
so your patch fits nicely.

| The original openpty patch is:
| 
| http://lists.nongnu.org/archive/html/axiom-developer/2006-11/msg00236.html
| 
| My version is below.

you would have gotten brownie points if you integrated


| The procedure to build this version of axiom is not too bad. Fink still
| seems to cause problems if the build machinery can see it, so I delete
| it from the path.

we can configure-detect that and set PATH accordingly.

| One caveat is that you need latex, makeindex, and

we can ask configure to return the absolute path so that there would
be no need to symlink from somewhere else.

| maybe tex for the axiom build to succeed, so I install them from fink,
| and link them to /usr/local/bin.
| 
| 
| $ PATH=/usr/bin:/bin:/usr/sbin:/sbin:/usr/X11R6/bin:/usr/local/bin
| $ ./configure
| $ make
| 
| The darn build takes like 6 hours on my dual G5 2.5 GHz Mac.

:-(

| I have some doubts about the patch, since I'm not really sure how to
| build configure, I'm not certain I did it right.

I very much appreciate your effort and patience with this.

| I don't have commit priviledges on the code base, and I don't want them
| until I'm more familiar with svn/svk, so could someone else please check
| in the patch?

Yes, I would like to work a little bit on it with you.

| Finally, this patch is against revision 326, because svk complains my
| axiom tree isn't local when I try to revise the patch against the
| current build-improvements head. I followed the svk instructions at:
| 
| http://wiki.axiom-developer.org/AxiomSilverBranch
| 
| but svk patch complains:
|   
| $ svk patch --view pty-again
| Target not local nor mirrored, unable to view patch.

>From the top my head, I don't see what that means.  I'll have to go back.


[...]

| === configure.ac.pamphlet
| ==================================================================
| --- configure.ac.pamphlet     (revision 326)
| +++ configure.ac.pamphlet     (patch pty-again level 1)
| @@ -676,8 +676,28 @@
|  AC_CHECK_HEADER([regex.h], [], [AC_MSG_ERROR([Axiom needs <regex.h>])])
|  @
|  
| +\subsubsection{openpty}
|  
| +[[clef]] and [[sman]] use ptys to communicate with [[AXIOMsys]]. The
| +existing code in silver doesn't work on MacOS X. We will try to build
| +with the Unix98 standard [[openpty]] function. We need to detect two
| +headers and a library. Linux has openpty defined in [[pty.h]], MacOS X
| +define it in [[util.h]]. FreeBSD is supposed to have a definition in
| +[[libutil.h]].

I like this.

| +<<headers>>=
| +AC_CHECK_HEADERS([pty.h util.h])
| +@
| +
| +On linux the openpty function is in [[libutil]]. We need to add that to
| +the list of libraries, at least for [[sman]] and [[clef]].

It would be good if you could integrate the suggestions here:

  http://lists.nongnu.org/archive/html/axiom-developer/2006-11/msg00687.html

In particular, we want to distinguish between the symbol openpty
(HAVE_OPENPTY) and its declaration (HAVE_OPENPTY_DECL).

[...]

| === src/lib/openpty.c.pamphlet
| ==================================================================
| --- src/lib/openpty.c.pamphlet        (revision 326)
| +++ src/lib/openpty.c.pamphlet        (patch pty-again level 1)
| @@ -10,17 +10,9 @@
|  \tableofcontents
|  \eject
|  \section{MAC OSX and BSD platform changes}
| -Since we have no other information we are adding the [[MACOSXplatform]] 
variable
| -to the list everywhere we find [[LINUXplatform]]. This may not be correct but
| -we have no way to know yet. We have also added the [[BSDplatform]] variable.
| -MAC OSX is some variant of BSD. These should probably be merged but we
| -cannot yet prove that.
| -<<mac osx platform change 1>>=
| -#if defined(SUN4OS5platform) ||defined(ALPHAplatform) || 
defined(HP10platform) || defined(LINUXplatform) || defined(MACOSXplatform) || 
defined(BSDplatform)
| -@
| -<<mac osx platform change 2>>=
| -#if defined(SUNplatform) || defined(HP9platform) || defined(LINUXplatform) 
|| defined(MACOSXplatform) || defined(BSDplatform)
| -@
| +We should really use autotools to check for Unix 98 pty support.
| +Before this is done below we hardcode information about each platform.

Yes.

| +
|  \section{License}
|  <<license>>=
|  /*
| @@ -71,6 +63,11 @@
|  #include <stropts.h>
|  #endif
|  
| +#ifdef HAVE_PTY_H
| +#include <pty.h>
| +#elifdef HAVE_UTIL_H

There is no standard C preprocessor directive #elifdef.  I believe you
wanted to say

    #elif defined(HAVE_UTIL_H)

Again, many thanks for your work and patience with this patch.

-- Gaby