From: Joel Brobecker <brobecker@adacore.com>
To: gdb-patches@sourceware.org
Subject: Re: [RFA/commit] configure: search waddstr only when building with TUI
Date: Fri, 03 Oct 2008 21:24:00 -0000 [thread overview]
Message-ID: <20081003212345.GS3665@adacore.com> (raw)
In-Reply-To: <20081003000916.GG26384@adacore.com>
[-- Attachment #1: Type: text/plain, Size: 2749 bytes --]
New version of the patch resent, with the following modifications:
- Remove the use of brobecker/<date> in the comment. Seems
like not everyone thinks it's useful. It doesn't bring much
value in specific this case anyways, so removed.
- Remove the quoting of parts of a C header from Tru64, to avoid
copyright issues.
The suggestions where made by Tom. Thank you, Tom, for the feedback.
> This one has been annoying us for a long time before we finally
> found the source of the problem. It happens only on Tru64, and
> the symptoms are a screwed terminal after GDB has started (missing
> carriage returns). The source of the problem is the following:
>
> * readline's configure determines that libtermcap is sufficient
> for its purposes, and thus includes termcap headers, etc.
>
> * GDB's configure has the following search:
> # For the TUI, we need enhanced curses functionality.
> [...]
> AC_SEARCH_LIBS(waddstr, [ncurses cursesX curses])
> This search is done even when TUI is not requested, and so
> causes GDB to be compiled with libcurses.
>
> On Tru64, unfortunately, the two libraries are not compatible.
> I provide a little more details in the patch itself.
>
> Ideally, we would want to adapt the build such that readline and
> GDB end up using the same library. But I haven't seen any other
> system but Tru64 where this is a problem, and fixing this seems
> a little tricky to coordinate.
>
> I side-stepped the issue and declared gdb with TUI broken on Tru64,
> and adjusted the configure to search for waddstr only when TUI was
> requested (enable_tui either "yes" or "auto"). On Tru64, we error
> out if TUI is requested, or disable it if "auto" was selected. The
> nice consequence of this change is that this should get rid of
> the dependency on libcurses on the platforms where its inclusion
> was useless...
>
> 2008-10-02 Joel Brobecker <brobecker@adacore.com>
>
> * configure.ac: On alpha-osf, error out if enable_tui is set to
> "yes", and set enable_tui to "no" if previously set to "auto".
> Check for waddstr only if TUI support was requested. Move the
> part of the configure script that updates various Makefile
> variables up, together with the check for waddstr.
> * configure: Regenerate.
>
> The patch looks big, but I'm just moving things around inside an if
> (except for the part were I error-out if TUI is requested on Tru64).
> Tested on x86-linux, no regression. Tested on tru64 as well, but I
> don't remember whether I ran the testsuite or not. I did verify
> that GDB is now linked with libtermcap rater than libcurses.
>
> Any objection? I'd like to commit this in about a week.
--
Joel
[-- Attachment #2: osf-tui.diff --]
[-- Type: text/plain, Size: 4779 bytes --]
diff -r 45c21f56c590 -r 9ac7e94abc9f configure.ac
--- a/configure.ac Thu Oct 02 16:18:27 2008 -0700
+++ b/configure.ac Fri Oct 03 14:19:59 2008 -0700
@@ -429,19 +429,69 @@ AC_SEARCH_LIBS(socketpair, socket)
# Link in zlib if we can. This allows us to read compressed debug sections.
AC_SEARCH_LIBS(zlibVersion, z, [AC_CHECK_HEADERS(zlib.h)])
-# For the TUI, we need enhanced curses functionality.
-#
-# FIXME: kettenis/20040905: We prefer ncurses over the vendor-supplied
-# curses library because the latter might not provide all the
-# functionality we need. However, this leads to problems on systems
-# where the linker searches /usr/local/lib, but the compiler doesn't
-# search /usr/local/include, if ncurses is installed in /usr/local. A
-# default installation of ncurses on alpha*-dec-osf* will lead to such
-# a situation.
-AC_SEARCH_LIBS(waddstr, [ncurses cursesX curses])
-
# On HP/UX we may need libxpdl for dlgetmodinfo (used by solib-pa64.c).
AC_SEARCH_LIBS(dlgetmodinfo, [dl xpdl])
+
+# On alpha-osf, it appears that libtermcap and libcurses are not compatible.
+# There is a very specific comment in /usr/include/curses.h explaining that
+# termcap routines built into libcurses must not be used.
+#
+# The symptoms we observed so far is GDB unexpectedly changing
+# the terminal settings when tgetent is called - this is particularly
+# visible as the output is missing carriage returns, and so rapidly
+# becomes very hard to read.
+#
+# The readline configure script has already decided that libtermcap
+# was enough for its purposes, and so decided to build readline using
+# libtermcap. Since the TUI mode requires curses, building GDB with
+# TUI enabled results in both libraries to be used at the same time,
+# which is not allowed. This basically means that GDB with TUI is
+# broken on alpha-osf.
+
+case $host_os in
+ alpha*-*-osf* )
+ if "$enable_tui" = "yes"; then
+ AC_MSG_ERROR([Building GDB with TUI mode is not supported on this host])
+ fi
+ if "$enable_tui" = "auto"; then
+ enable_tui=no
+ fi
+ ;;
+esac
+
+# Check whether we should enable the TUI, but only do so if we really
+# can.
+if test x"$enable_tui" != xno; then
+ if test -d $srcdir/tui; then
+ # For the TUI, we need enhanced curses functionality.
+ #
+ # FIXME: kettenis/20040905: We prefer ncurses over the vendor-supplied
+ # curses library because the latter might not provide all the
+ # functionality we need. However, this leads to problems on systems
+ # where the linker searches /usr/local/lib, but the compiler doesn't
+ # search /usr/local/include, if ncurses is installed in /usr/local. A
+ # default installation of ncurses on alpha*-dec-osf* will lead to such
+ # a situation.
+ AC_SEARCH_LIBS(waddstr, [ncurses cursesX curses])
+
+ if test "$ac_cv_search_waddstr" != no; then
+ CONFIG_OBS="$CONFIG_OBS \$(SUBDIR_TUI_OBS)"
+ CONFIG_DEPS="$CONFIG_DEPS \$(SUBDIR_TUI_DEPS)"
+ CONFIG_SRCS="$CONFIG_SRCS \$(SUBDIR_TUI_SRCS)"
+ ENABLE_CFLAGS="$ENABLE_CFLAGS \$(SUBDIR_TUI_CFLAGS)"
+ CONFIG_ALL="${CONFIG_ALL} all-tui"
+ CONFIG_CLEAN="${CONFIG_CLEAN} clean-tui"
+ CONFIG_INSTALL="${CONFIG_INSTALL} install-tui"
+ CONFIG_UNINSTALL="${CONFIG_UNINSTALL} uninstall-tui"
+ else
+ if test x"$enable_tui" = xyes; then
+ AC_MSG_ERROR([no enhanced curses library found; disable TUI])
+ else
+ AC_MSG_WARN([no enhanced curses library found; disabling TUI])
+ fi
+ fi
+ fi
+fi
# Since GDB uses Readline, we need termcap functionality. In many
# cases this will be provided by the curses library, but some systems
@@ -1748,29 +1798,6 @@ AC_SUBST(GDBTK_SRC_DIR)
AC_PATH_X
-# Check whether we should enable the TUI, but only do so if we really
-# can.
-if test x"$enable_tui" != xno; then
- if test -d $srcdir/tui; then
- if test "$ac_cv_search_waddstr" != no; then
- CONFIG_OBS="$CONFIG_OBS \$(SUBDIR_TUI_OBS)"
- CONFIG_DEPS="$CONFIG_DEPS \$(SUBDIR_TUI_DEPS)"
- CONFIG_SRCS="$CONFIG_SRCS \$(SUBDIR_TUI_SRCS)"
- ENABLE_CFLAGS="$ENABLE_CFLAGS \$(SUBDIR_TUI_CFLAGS)"
- CONFIG_ALL="${CONFIG_ALL} all-tui"
- CONFIG_CLEAN="${CONFIG_CLEAN} clean-tui"
- CONFIG_INSTALL="${CONFIG_INSTALL} install-tui"
- CONFIG_UNINSTALL="${CONFIG_UNINSTALL} uninstall-tui"
- else
- if test x"$enable_tui" = xyes; then
- AC_MSG_ERROR([no enhanced curses library found; disable TUI])
- else
- AC_MSG_WARN([no enhanced curses library found; disabling TUI])
- fi
- fi
- fi
-fi
-
# Unlike the sim directory, whether a simulator is linked is controlled by
# presence of a gdb_sim definition in the target configure.tgt entry.
# This code just checks for a few cases where we'd like to ignore those
next prev parent reply other threads:[~2008-10-03 21:24 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-10-03 0:10 Joel Brobecker
2008-10-03 21:24 ` Joel Brobecker [this message]
2008-10-22 20:25 ` Joel Brobecker
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20081003212345.GS3665@adacore.com \
--to=brobecker@adacore.com \
--cc=gdb-patches@sourceware.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox