pspp-builds.git
14 years agoSet final element in covariance matrix. Use accessor functions for linreg struct... origin/covariance fc11-i386-build90 fc11-x64-build87 lenny-x64-build105 sid-i386-build157
Jason H Stover [Fri, 5 Feb 2010 20:21:17 +0000 (15:21 -0500)]
Set final element in covariance matrix. Use accessor functions for linreg struct. Initialize total sums of squares.

14 years agouse sizeof double to allocate means fc11-i386-build89 fc11-x64-build86 lenny-x64-build104 sid-i386-build156
Jason H Stover [Thu, 4 Feb 2010 23:01:45 +0000 (18:01 -0500)]
use sizeof double to allocate means

14 years agoCompute the independent variable means fc11-i386-build88 fc11-x64-build85 lenny-x64-build103 sid-i386-build155
Jason H Stover [Thu, 4 Feb 2010 22:26:54 +0000 (17:26 -0500)]
Compute the independent variable means

14 years agoallocate space to store independent variable list fc11-i386-build86 fc11-x64-build83 lenny-x64-build101 sid-i386-build153
Jason H Stover [Sat, 30 Jan 2010 04:15:07 +0000 (23:15 -0500)]
allocate space to store independent variable list

14 years agofixed counting of valid cases in fill_covariance fc11-i386-build84 fc11-x64-build81 lenny-x64-build99 sid-i386-build151
Jason H Stover [Fri, 29 Jan 2010 19:25:09 +0000 (14:25 -0500)]
fixed counting of valid cases in fill_covariance

14 years agoFree linreg struct if no valid data; fc11-i386-build83 fc11-x64-build80 lenny-x64-build98 sid-i386-build150
Jason H Stover [Fri, 29 Jan 2010 19:01:58 +0000 (14:01 -0500)]
Free linreg struct if no valid data;
no not attempt to save residuals or predicted values in this case

14 years agoSet categoricals to NULL in one-pass covariance; avoid freeing NULL in categoricals_d... fc11-i386-build81 fc11-x64-build78 sid-i386-build148
Jason H Stover [Fri, 22 Jan 2010 21:48:04 +0000 (16:48 -0500)]
Set categoricals to NULL in one-pass covariance; avoid freeing NULL in categoricals_destroy

14 years agofill_covariance now returns a value fc11-i386-build79 fc11-x64-build76 sid-i386-build146
Jason H Stover [Thu, 21 Jan 2010 22:25:27 +0000 (17:25 -0500)]
fill_covariance now returns a value

14 years agoRewrote linreg.[ch], regression.q and glm.q to use the new fc11-i386-build71 fc11-x64-build68 lenny-x64-build92 sid-i386-build138
Jason H Stover [Fri, 15 Jan 2010 20:17:54 +0000 (15:17 -0500)]
Rewrote linreg.[ch], regression.q and glm.q to use the new
covariance.[ch]. Dropped covariance-matrix.[ch], design-matrix.[ch]
and coefficient.[ch].

14 years agoCall covariance_accumulate_pass2 fc11-i386-build39 fc11-x64-build36 lenny-x64-build60 sid-i386-build108
Jason H Stover [Wed, 18 Nov 2009 21:51:33 +0000 (16:51 -0500)]
Call covariance_accumulate_pass2

14 years agoFix separation of categorical and numeric variables. fc11-i386-build37 fc11-x64-build34 lenny-x64-build58 sid-i386-build106
Jason H Stover [Tue, 17 Nov 2009 21:48:45 +0000 (16:48 -0500)]
Fix separation of categorical and numeric variables.
Enforce numeric type for dependent variables.
Name the covariance struct returned by covariance_2pass_create as cov.

14 years agoWrap operations which index into reverse_variable_map fc11-i386-build30 fc11-x64-build27 lenny-x64-build51 sid-i386-build99
John Darrington [Sat, 31 Oct 2009 08:02:32 +0000 (09:02 +0100)]
Wrap operations which index into reverse_variable_map

New private function reverse_variable_lookup.  This
function wraps operations which index into the array,
and includes checks to prevent indexing outside the
bounds of the array.

14 years agoDrop first category of each variable from covariance matrix.
John Darrington [Sat, 31 Oct 2009 07:42:25 +0000 (08:42 +0100)]
Drop first category of each variable from covariance matrix.

Apparently it's not appropriate for the covariance matrix to
have a row/column for every categorical value.  One value per
categorical variable should be excluded.  This change drops
the first category encountered.

14 years agoRemove unused function categoricals_index
John Darrington [Sat, 31 Oct 2009 07:22:00 +0000 (08:22 +0100)]
Remove unused function categoricals_index

14 years agoPopulate the moment matrices with the categorical entries
John Darrington [Sat, 31 Oct 2009 07:10:35 +0000 (08:10 +0100)]
Populate the moment matrices with the categorical entries

14 years agoKeep track of the number of variables involved in the categoricals.
John Darrington [Sat, 31 Oct 2009 07:04:32 +0000 (08:04 +0100)]
Keep track of the number of variables involved in the categoricals.

Added a counter for the number of variables which actually contribute
to categories.  Due to the possibility of missing values, this may
not be equal to the number of variables passed to the contructor.

14 years agoDon't crash if all categorical variables are empty
John Darrington [Sat, 31 Oct 2009 06:04:25 +0000 (07:04 +0100)]
Don't crash if all categorical variables are empty

14 years agoNew functions to return the sum and weights associated with categorical variables fc11-i386-build28 fc11-x64-build25 lenny-x64-build49 sid-i386-build97
John Darrington [Sat, 24 Oct 2009 14:20:10 +0000 (16:20 +0200)]
New functions to return the sum and weights associated with categorical variables

14 years agoDon't update categorical variables whose values are missing
John Darrington [Sat, 24 Oct 2009 14:14:59 +0000 (16:14 +0200)]
Don't update categorical variables whose values are missing

14 years agoRemove some unnecessary #include directives
John Darrington [Sat, 24 Oct 2009 08:27:54 +0000 (10:27 +0200)]
Remove some unnecessary #include directives

14 years agoCall categoricals_done at start of second pass. sid-i386-build96
John Darrington [Sat, 24 Oct 2009 07:26:21 +0000 (09:26 +0200)]
Call categoricals_done at start of second pass.

14 years agoNew functions is_missing and get_val
John Darrington [Sat, 24 Oct 2009 07:02:54 +0000 (09:02 +0200)]
New functions is_missing and get_val

Abstract the actions of obtaining the numeric value of a variable
and its missing status, so that it can be more easily applied to
categorical variables and linear variables alike.

14 years agoUse cov->dim instead of cov->n_vars where appropriate
John Darrington [Sat, 24 Oct 2009 06:35:43 +0000 (08:35 +0200)]
Use cov->dim instead of cov->n_vars where appropriate

14 years agoDestroy the categorical variables when the covariance matrix is destroyed
John Darrington [Sat, 24 Oct 2009 06:32:48 +0000 (08:32 +0200)]
Destroy the categorical variables when the covariance matrix is destroyed

14 years agoMove categorical variable into the var_params struct
John Darrington [Sat, 24 Oct 2009 06:29:45 +0000 (08:29 +0200)]
Move categorical variable into the var_params struct

14 years agoResize the moment matrices after the first pass
John Darrington [Fri, 23 Oct 2009 17:16:31 +0000 (19:16 +0200)]
Resize the moment matrices after the first pass

14 years agoAllow dump function to work with numeric variables
John Darrington [Tue, 20 Oct 2009 17:27:54 +0000 (19:27 +0200)]
Allow dump function to work with numeric variables

14 years agoRework glm.q to use new covariance routines fc11-i386-build27 fc11-x64-build24 lenny-x64-build48 sid-i386-build95
Jason H Stover [Fri, 23 Oct 2009 20:37:11 +0000 (16:37 -0400)]
Rework glm.q to use new covariance routines

14 years agoFix order of pool_calloc arguments. fc11-i386-build26 fc11-x64-build22 lenny-x64-build47 sid-i386-build94
John Darrington [Sun, 18 Oct 2009 19:57:42 +0000 (21:57 +0200)]
Fix order of pool_calloc arguments.

The second argument is the number of elements, the third is the size.  If this
is inadvertently reversed, then it fails when the number of elements is zero.

14 years agoAdd reverse map entries to categoricals.
John Darrington [Sun, 18 Oct 2009 19:46:46 +0000 (21:46 +0200)]
Add reverse map entries to categoricals.

Allow categorical values and variables to be efficiently
indexed by the non-zero binary vector entry to which they
map.

14 years agoNew module src/math/categoricals .
John Darrington [Sat, 17 Oct 2009 17:26:04 +0000 (19:26 +0200)]
New module src/math/categoricals .

Created a new object to keep track of categorical variables, and their
values.   This is similar to src/data/categorys.c but is created on
demand, and doesn't use any global state.

14 years agoAdd a two pass algorithm to calculate covariance matrices. fc11-i386-build23 lenny-x64-build44 sid-i386-build91
John Darrington [Sat, 17 Oct 2009 09:12:56 +0000 (11:12 +0200)]
Add a two pass algorithm to calculate covariance matrices.

The single pass algorithm, although faster, is numerically unstable.
Furthermore, at can't reasonably be achieved when categorical variables
are involved.    This change implements a two pass algorithm and allows
the caller to choose which algorithm (s)he prefers.

14 years agoNew function covariance_2pass_create fbsd72-i386-build31 fc11-i386-build17 fc11-x64-build18 lenny-x64-build39 sid-i386-build85
Jason H Stover [Thu, 8 Oct 2009 21:19:11 +0000 (17:19 -0400)]
New function covariance_2pass_create

14 years agoUpdate NEWS file fbsd72-i386-build29 fc11-i386-build15 fc11-x64-build16 lenny-x64-build37 sid-i386-build83
John Darrington [Tue, 6 Oct 2009 16:36:40 +0000 (18:36 +0200)]
Update NEWS file

14 years agoNew function check_interactions to specify all possible interactions fbsd72-i386-build28 fc11-i386-build14 fc11-x64-build15 lenny-x64-build36 sid-i386-build82
Jason H Stover [Tue, 6 Oct 2009 14:33:03 +0000 (10:33 -0400)]
New function check_interactions to specify all possible interactions
in the absence of the DESIGN subcommand.

Remove calls to covariance_matrix routines as preparation to rewrite
src/math/covariance-matrix.c.

14 years agohmap-test: Skip test instead of reporting success on GCC 4.3. fbsd72-i386-build27 fc11-i386-build13 fc11-x64-build14 lenny-x64-build35 sid-i386-build81
Ben Pfaff [Mon, 5 Oct 2009 03:37:10 +0000 (20:37 -0700)]
hmap-test: Skip test instead of reporting success on GCC 4.3.

Commit 1dfc6b1 "hmap-test: Disable tests that GCC 4.3 miscompiles."
disabled some of the hmap-tests, the ones that GCC 4.3 miscompiles.
However, it still reported "success" if the remaining tests succeeded.
John Darrington suggested that reporting that the test was skipped was
more accurate, so this commit does so.

14 years agoUpgrade gtk-builder-convert to newer version that adds useful warnings. fbsd72-i386-build26 fc11-i386-build12 fc11-x64-build13 lenny-x64-build34 sid-i386-build80
Ben Pfaff [Mon, 5 Oct 2009 03:27:30 +0000 (20:27 -0700)]
Upgrade gtk-builder-convert to newer version that adds useful warnings.

The notes on "Migrating from libglade to GtkBuilder" in the GTK+
Reference Manual at, e.g.
http://library.gnome.org/devel/gtk/unstable/gtk-migrating-GtkBuilder.html
say that "While libglade can often tolerate multiple widgets having the
same id in a glade file, GtkBuilder will not accept duplicate object ids."
The version of gtk-builder-convert that we had previously didn't warn about
duplicate ids; the new version checked in by this commit does.

14 years agoCanonicalise orthography of Big/Little Endian fbsd72-i386-build25 fc11-i386-build11 fc11-x64-build12 lenny-x64-build33 sid-i386-build79
John Darrington [Sun, 4 Oct 2009 16:58:39 +0000 (18:58 +0200)]
Canonicalise orthography of Big/Little Endian

14 years agoRemove duplicate entry from command.def
John Darrington [Sun, 4 Oct 2009 16:52:15 +0000 (18:52 +0200)]
Remove duplicate entry from command.def

14 years agoUpdate INSTALL file
John Darrington [Sun, 4 Oct 2009 15:46:14 +0000 (17:46 +0200)]
Update INSTALL file

14 years agoCORRELATIONS: Properly deal with case weights. fbsd72-i386-build24 fc11-i386-build10 fc11-x64-build11 lenny-x64-build32 sid-i386-build78
John Darrington [Sun, 4 Oct 2009 14:51:57 +0000 (16:51 +0200)]
CORRELATIONS: Properly deal with case weights.

Corrected a bug in the weighted covariance calculations,
and add a test to check that weights are properly
handled.

14 years agoCORRELATIONS: Fix bug parsing /STATISTICS=ALL
John Darrington [Sun, 4 Oct 2009 14:25:04 +0000 (16:25 +0200)]
CORRELATIONS: Fix bug parsing /STATISTICS=ALL

14 years agoCORRELATIONS: Documented the /STATISTICS subcommand
John Darrington [Sun, 4 Oct 2009 14:06:14 +0000 (16:06 +0200)]
CORRELATIONS: Documented the /STATISTICS subcommand

14 years agoCORRELATIONS: Added support for /STATISTICS=XPROD
John Darrington [Sun, 4 Oct 2009 13:54:00 +0000 (15:54 +0200)]
CORRELATIONS: Added support for /STATISTICS=XPROD

14 years agoCORRELATIONS: Added support for the STATISTICS subcommand
John Darrington [Sun, 4 Oct 2009 12:59:34 +0000 (14:59 +0200)]
CORRELATIONS: Added support for the STATISTICS subcommand

14 years agoAdd a manual entry for the CORRELATIONS command
John Darrington [Sat, 3 Oct 2009 20:46:11 +0000 (22:46 +0200)]
Add a manual entry for the CORRELATIONS command

14 years agoFirst working version of CORRELATIONS.
John Darrington [Sat, 3 Oct 2009 19:52:01 +0000 (21:52 +0200)]
First working version of CORRELATIONS.

This commit includes a new module, src/math/covariance.[ch],
which may eventually replace src/math/covariance-matrix.[ch]

14 years agoInitial framework for CORRELATIONS command
John Darrington [Sat, 3 Oct 2009 06:42:44 +0000 (08:42 +0200)]
Initial framework for CORRELATIONS command

14 years agoFix numerous warnings about an unknown attribute rpl_malloc on mingw32. fbsd72-i386-build23 fc11-i386-build9 fc11-x64-build10 lenny-x64-build31 sid-i386-build77
Ben Pfaff [Thu, 1 Oct 2009 04:50:13 +0000 (21:50 -0700)]
Fix numerous warnings about an unknown attribute rpl_malloc on mingw32.

On mingw32, Gnulib replaces malloc by defining it as a macro that expands
to rpl_malloc.  Since MALLOC_LIKE used malloc as an identifier, even though
it didn't refer to the malloc function per se, this instance of malloc
also expanded to rpl_malloc, confusing GCC.

GCC also accepts __malloc__ here, so this commit uses that spelling
instead, which avoids the warning.

14 years agoUpgrade to latest Gnulib version and fix resulting FreeBSD build failures. fbsd72-i386-build22 fc11-i386-build8 fc11-x64-build9 lenny-x64-build30 sid-i386-build76
Ben Pfaff [Thu, 1 Oct 2009 04:45:47 +0000 (21:45 -0700)]
Upgrade to latest Gnulib version and fix resulting FreeBSD build failures.

Upgrading the Gnulib version exposed a couple of failures to #include
<config.h> at the very top of a source file via build failures on
FreeBSD 7.2-RELEASE, fixed by this commit.

This probably also fixes a similar OpenBSD issue reported by Jason Stover
at http://lists.gnu.org/archive/html/pspp-dev/2009-09/msg00095.html

14 years agoREADME.Git: Fix syntax of suggested commands. fbsd72-i386-build17 fbsd72-i386-build18 fbsd72-i386-build19 fbsd72-i386-build21 fc11-i386-build5 fc11-i386-build6 fc11-i386-build7 fc11-x64-build6 fc11-x64-build7 fc11-x64-build8 lenny-x64-build27 lenny-x64-build28 lenny-x64-build29 sid-i386-build70 sid-i386-build71 sid-i386-build72 sid-i386-build75
Ben Pfaff [Thu, 1 Oct 2009 03:40:22 +0000 (20:40 -0700)]
README.Git: Fix syntax of suggested commands.

14 years agoINSTALL: Mention that GNU make is required.
Ben Pfaff [Thu, 1 Oct 2009 03:38:26 +0000 (20:38 -0700)]
INSTALL: Mention that GNU make is required.

In my experience with the autobuilder, in particular the system "make"
program on FreeBSD 7.2-RELEASE will not work.

14 years agoRewrite README.Git to mention the Gnulib commit number and explain better. fbsd72-i386-build16 fc11-i386-build4 fc11-x64-build5 lenny-x64-build26
Ben Pfaff [Thu, 1 Oct 2009 03:33:34 +0000 (20:33 -0700)]
Rewrite README.Git to mention the Gnulib commit number and explain better.

This commit has three purposes.  First, I think that the new version of this
file explains the bootstrapping process a little better.  Second, I want
to mention a Gnulib commit number that I have been using, so that other
people can easily figure out a version of Gnulib that will work.  Third,
I want the PSPP autobuilder to be able to automatically pick out a working
Gnulib commit.

14 years agoperl-module: Fix build on FreeBSD 7.2 fbsd72-i386-build15 fc11-i386-build3 fc11-x64-build4 lenny-x64-build25 sid-i386-build69
Ben Pfaff [Wed, 30 Sep 2009 04:47:09 +0000 (21:47 -0700)]
perl-module: Fix build on FreeBSD 7.2

config.h must be included first, otherwise the build fails on FreeBSD 7.2
in gl/string.h at the first instance of "restrict", which <config.h>
#defines as empty.

14 years agoDon't use nonportable "mktemp" utility in perl-module/automake.mk.
Ben Pfaff [Wed, 30 Sep 2009 04:33:34 +0000 (21:33 -0700)]
Don't use nonportable "mktemp" utility in perl-module/automake.mk.

In porting to FreeBSD, I noticed that perl-module/automake.mk tries to use
the "mktemp" utility.  FreeBSD happens to have this utility, but it
requires an argument, and at any rate it is nonportable and we can avoid
using it here, so do so.

14 years agoAdded a new target "programs" to the makefile. fbsd72-i386-build11 fbsd72-i386-build12 fbsd72-i386-build13 fbsd72-i386-build14 fbsd72-i386-build3 fbsd72-i386-build4 fbsd72-i386-build5 fbsd72-i386-build6 fbsd72-i386-build7 fbsd72-i386-build8 fc11-i386-build1 fc11-i386-build2 fc11-x64-build1 fc11-x64-build2 fc11-x64-build3 lenny-x64-build22 lenny-x64-build23 lenny-x64-build24 sid-i386-build67 sid-i386-build68
John Darrington [Mon, 28 Sep 2009 18:27:04 +0000 (20:27 +0200)]
Added a new target "programs" to the makefile.

When a source file is changed, the "all" target (correctly) rebuilds
the po/pspp.pot file and remerges the PO files for all supported translations,
and then builds .gmo files for each of them.   Whilst this is correct behaviour,
it will become an unmitigated nuisance for developers as more and more
translations are added.

This change adds a new target called "programs" which builds the binaries, but
not the ancilliary targets. This target is more convienient for working on
changes to the source code, when translations are not of interest.

14 years agoCorrect printf conversion specifiers for size_t types. fc11-i386-build0 lenny-x64-build21 sid-i386-build66
John Darrington [Mon, 28 Sep 2009 18:12:54 +0000 (20:12 +0200)]
Correct printf conversion specifiers for size_t types.

14 years agoAdd necessary #include stddef.h directive build52 lenny-x64-build16 lenny-x64-build17 lenny-x64-build18 lenny-x64-build19 lenny-x64-build20 sid-i386-build59 sid-i386-build60 sid-i386-build61 sid-i386-build62 sid-i386-build63 sid-i386-build64 sid-i386-build65
John Darrington [Sun, 27 Sep 2009 14:11:25 +0000 (16:11 +0200)]
Add necessary #include stddef.h directive

14 years agoperl-module: Additional fixes for Perl version number on PSPP autobuilder. origin/perl build3 build4 build49 build50 build51
Ben Pfaff [Sat, 26 Sep 2009 04:14:31 +0000 (21:14 -0700)]
perl-module: Additional fixes for Perl version number on PSPP autobuilder.

Commit 30f21302 "perl-module: Tolerate minor PSPP version number
differences" tried to improve the behavior of the PSPP Perl module when
the PSPP version number includes a suffix such as -build40.  But it didn't
entirely work, since it overlooked some uses of $(VERSION) in
perl-module/automake.mk.  This commit fixes those too.

14 years agoFix missing @clicksequence problem for building Docbook output also. build1 build48
Ben Pfaff [Sat, 26 Sep 2009 03:38:07 +0000 (20:38 -0700)]
Fix missing @clicksequence problem for building Docbook output also.

Commit 770f21f5 "Fix missing @clicksequence problem with older Texinfo
versions." fixed the problem of missing @clicksequence for building the
main PSPP documentation, partly by adding to $(AM_MAKEINFOFLAGS).  However,
$(AM_MAKEINFOFLAGS) wasn't used in building the Docbook output for use by
Yelp as the GUI help.  This commit adds $(AM_MAKEINFOFLAGS) to that
command, fixing the problem.

14 years agoperl-module: Tolerate minor PSPP version number differences. build0 build41 build42 build43 build44 build45 build46 build47
Ben Pfaff [Fri, 25 Sep 2009 04:38:09 +0000 (21:38 -0700)]
perl-module: Tolerate minor PSPP version number differences.

The PSPP autobuilder that I'm working on appends a build number to the PSPP
version number, e.g. "0.7.2-build40".  The Perl module has two problems
with this:

  1. Perl won't parse version numbers that contain anything other than
     digits and periods, so "-build" causes an error.  So this commit
     drops everything from the hyphen onward from the PSPP version number
     passed along to Perl.

  2. The Perl module itself does a string comparison operation between the
     number it expects and the number that PSPP reports.  When we drop
     the hyphen onward, the comparison fails.  So this commit also relaxes
     the comparison to one that does pass.

14 years agoMerge commit 'origin/stable' build40
John Darrington [Thu, 24 Sep 2009 18:43:42 +0000 (20:43 +0200)]
Merge commit 'origin/stable'

Conflicts:

AUTHORS
NEWS
configure.ac
lib/gtksheet/gtkitementry.c
po/LINGUAS
po/nl.po
src/data/file-name.c
src/data/sys-file-writer.c
src/language/stats/crosstabs.q
src/language/stats/examine.q
src/language/stats/t-test.q
src/output/charts/box-whisker.c
src/output/charts/plot-hist.c
src/ui/gui/output-viewer.c
src/ui/gui/psppire.c
src/ui/gui/psppire.h
src/ui/terminal/main.c

14 years agoMake GtkItemEntry compatible with GTK+ 2.17.4 and later.
Ben Pfaff [Thu, 24 Sep 2009 02:45:56 +0000 (19:45 -0700)]
Make GtkItemEntry compatible with GTK+ 2.17.4 and later.

GTK+ 2.17.4 breaks GtkEntry's buffer into a separate class named
GtkEntryBuffer and eliminates the associated members from GtkEntry.
The GtkItemEntry code in lib/gtksheet attempted to use some of those
members directly, which no longer works, so this commit changes that
code to use public function interfaces instead.  Some of these public
function interfaces were introduced in GTK+ 2.17.4, so those pieces
are conditionalized on the version of GTK+ detected at build time.

This commit makes PSPPIRE compiled with GTK+ before 2.17.4 work properly
only when used with GTK+ before 2.17.4, and similarly when PSPPIRE is
compiled with GTK+ 2.17.4 or later.  There may be a way to avoid that, but
it is not entirely obvious.

With feedback from John Darrington.

14 years agoRename README.CVS to README.Git and update to refer to Git instead of CVS. build37 build38 build39
Ben Pfaff [Wed, 23 Sep 2009 14:40:26 +0000 (07:40 -0700)]
Rename README.CVS to README.Git and update to refer to Git instead of CVS.

14 years agoFix "make install" with --enable-relocatable.
Ben Pfaff [Wed, 23 Sep 2009 05:19:17 +0000 (22:19 -0700)]
Fix "make install" with --enable-relocatable.

When PSPP added support for --enable-relocatable back in 2007, PSPP did
not install any shared libraries.  Now it does, which requires us to
set RELOCATABLE_LIBRARY_PATH in configure.ac.  Otherwise, the installed
binaries will not work automatically but require some manual user
intervention.

Also set RELOCATABLE_STRIP as recommended by the gnulib documentation.

14 years agohmap-test: Disable tests that GCC 4.3 miscompiles.
Ben Pfaff [Mon, 21 Sep 2009 04:38:12 +0000 (21:38 -0700)]
hmap-test: Disable tests that GCC 4.3 miscompiles.

Commit ac58b3c51 "Avoid bug in GCC 4.3 in hmap-test." attempted to work
around a bug in GCC 4.3 that was causing test failures in hmap-test.
Unfortunately, the problem has started resurfacing again, probably due to
changes elsewhere, so this commit simply disables the tests in question
when PSPP is built with GCC 4.3.

14 years agoFix "make check" with --enable-relocatable.
Ben Pfaff [Sun, 20 Sep 2009 21:04:28 +0000 (14:04 -0700)]
Fix "make check" with --enable-relocatable.

When --enable-relocatable was passed to "configure", "make check" failed
while building datasheet-test:

tests/data/datasheet-test.c: In function 'main':
tests/data/datasheet-test.c:880: error: 'INSTALLDIR' undeclared (first use in this function)

Line 880 of that file is:

set_program_name (argv[0]);

which is a little mysterious until one realizes that --enable-relocatable
defines set_program_name() as a macro that expands to a value that includes
INSTALLDIR.

So this commit fixes the problem by making sure that INSTALLDIR is defined
as part of CPPFLAGS for all compiles.

14 years agoDon't pad variable labels written to system files.
Ben Pfaff [Thu, 17 Sep 2009 03:19:34 +0000 (20:19 -0700)]
Don't pad variable labels written to system files.

Previously, the code to write variable labels to system files accidentally
padded out the variable labels on the right with spaces to a multiple of 4
bytes in length.  With this commit, variable labels will no longer be
padded.

This fixes a genuine bug for variable labels longer than 252 bytes, which
previously were padded out to 256 bytes.  Variable labels are limited to
a maximum of 255 bytes, so PSPP refused to read such files.

Reported-by: Robert Westlund <rob@kinesissurvey.com>
14 years agoFREQUENCIES: Make frequency table headings easier to translate.
Ben Pfaff [Wed, 16 Sep 2009 04:12:34 +0000 (21:12 -0700)]
FREQUENCIES: Make frequency table headings easier to translate.

The headings for frequency tables were broken up into individual words,
which made them hard to translate since in some languages the words needed
to be in the opposite order.

Thanks to Harry Thijssen <pspp@sjpaes.nl> for pointing out the problem.

14 years agoMerge commit 'origin/stable' into stable
Ben Pfaff [Tue, 15 Sep 2009 03:41:35 +0000 (20:41 -0700)]
Merge commit 'origin/stable' into stable

14 years agoFix tests broken by the new table widths, caused by previous change
John Darrington [Sat, 12 Sep 2009 14:35:17 +0000 (16:35 +0200)]
Fix tests broken by the new table widths, caused by previous change

14 years agoCanonicalise some printed strings, to make translators' lives easier
John Darrington [Sat, 12 Sep 2009 09:15:47 +0000 (11:15 +0200)]
Canonicalise some printed strings, to make translators' lives easier

14 years agoRelease PSPP 0.6.2-pre6. v0.6.2-pre6
Ben Pfaff [Wed, 9 Sep 2009 04:43:01 +0000 (21:43 -0700)]
Release PSPP 0.6.2-pre6.

14 years agoAdd Brazilian Portuguese translation.
Ben Pfaff [Wed, 9 Sep 2009 03:59:04 +0000 (20:59 -0700)]
Add Brazilian Portuguese translation.

Thanks to Michel Boaventura <michel@cecaps.ufmg.br> and the coordinators
at translationproject.org.

14 years agoT-TEST: Also fix the case where the correlation is slightly less than -1.0.
Ben Pfaff [Fri, 4 Sep 2009 16:05:08 +0000 (09:05 -0700)]
T-TEST: Also fix the case where the correlation is slightly less than -1.0.

Thanks to John Darrington for pointing out the omission.

14 years agoUpdate to latest nl.po.
Ben Pfaff [Fri, 4 Sep 2009 03:31:38 +0000 (20:31 -0700)]
Update to latest nl.po.

Thanks to unknown-1 and the coordinators at translationproject.org.

14 years agoFix missing @clicksequence problem with older Texinfo versions. master
Ben Pfaff [Tue, 4 Aug 2009 13:01:21 +0000 (06:01 -0700)]
Fix missing @clicksequence problem with older Texinfo versions.

The @clicksequence command was introduced in Texinfo 4.12, which a lot of
users do not yet have.  This commit fixes the problem by detecting whether
makeinfo supports @clicksequence and, if not, defining a macro for it as
a substitute.

We also need support for @clicksequence in texinfo.tex.  It is possible to
detect whether @clicksequence is defined using TeX code and, if not,
define a TeX macro for it.  However, I'm not sure how to do that.  So this
commit instead adds a known-good texinfo.tex to the source tree, which
should have the same effect.

14 years agoT-TEST: Avoid NaN in paired-sample correlation significance calcuation.
Ben Pfaff [Fri, 4 Sep 2009 03:09:36 +0000 (20:09 -0700)]
T-TEST: Avoid NaN in paired-sample correlation significance calcuation.

Sometimes the correlation comes out just above 1.0 due to inaccuracy of
machine calculation, which causes sqrt(1-corr**2) to yield NaN.  This
commit forces the correlation into the valid range, fixing the problem.

Thanks to Matej Cepl <mcepl@redhat.com> for reporting the problem.
Thanks to John Darrington for suggesting the fix.

14 years agoNew functions glm_custom_design and parse_interactions.
Jason Stover [Wed, 2 Sep 2009 01:21:02 +0000 (21:21 -0400)]
New functions glm_custom_design and parse_interactions.

14 years agoRemove i18n related configuration items which are no longer used
John Darrington [Tue, 1 Sep 2009 20:17:56 +0000 (22:17 +0200)]
Remove i18n related configuration items which are no longer used

14 years agoReplace "all-hook" with "all-local".
John Darrington [Mon, 31 Aug 2009 18:32:00 +0000 (20:32 +0200)]
Replace "all-hook" with "all-local".

all-hook isn't a target recognised by automake, so use all-local instead,
and avoid it clashing with the all-local in perl-module

14 years agoAdd the all_q_sources to the dependencies of po/pspp.pot
John Darrington [Mon, 31 Aug 2009 18:15:26 +0000 (20:15 +0200)]
Add the all_q_sources to the dependencies of po/pspp.pot

14 years agoRewrite automake rules for internationalisation files.
John Darrington [Mon, 31 Aug 2009 16:25:23 +0000 (18:25 +0200)]
Rewrite automake rules for internationalisation files.

I got fed up with battling against the Makefile which
gettext provides.  So I rolled my own instead.  This change provides
the following advantages over previous situation:

* It doesn't use recursive calls to $(MAKE).

* Merged and unmerged po files are properly seperated into
  the buildir and sourcedir respectively.  Consequently,
  po files get updated only when we really want them to be,
  and not (potentially) on every commit.

* The .pot file (and consequently the .po and .gmo files) now
  properly depend upon all the distributed source files.  There
  is no need to run "make -f Smake" to make it aware of new files.

* There's no more need of "timestamp" files.

* Potentially it's easier to add per-source-file options
  when generating the .pot file. (and thus avoid warnings about
  unrecognised extentions.

14 years agoCorrect typo in printed string
John Darrington [Sun, 30 Aug 2009 14:52:53 +0000 (16:52 +0200)]
Correct typo in printed string

14 years agoUpdate to most recent Gnulib
John Darrington [Sun, 30 Aug 2009 14:51:16 +0000 (16:51 +0200)]
Update to most recent Gnulib

14 years agoFix make distcheck
John Darrington [Sun, 30 Aug 2009 14:16:38 +0000 (16:16 +0200)]
Fix make distcheck

14 years agoDon't override label for Save As menuitem
John Darrington [Sun, 16 Aug 2009 05:27:01 +0000 (07:27 +0200)]
Don't override label for Save As menuitem

14 years agoCruft removal
Jason Stover [Mon, 3 Aug 2009 15:26:56 +0000 (11:26 -0400)]
Cruft removal

14 years agoUse u8_cpy instead of memcpy.
John Darrington [Thu, 30 Jul 2009 04:21:38 +0000 (06:21 +0200)]
Use u8_cpy instead of memcpy.

Use u8_cpy instead of memcpy since the
former provides better type checking.

14 years agoMerged changes from branch interaction-review
Jason Stover [Wed, 29 Jul 2009 01:32:31 +0000 (21:32 -0400)]
Merged changes from branch interaction-review

14 years agoRemove commented code
John Darrington [Tue, 28 Jul 2009 11:02:06 +0000 (13:02 +0200)]
Remove commented code

14 years agoremove unused return value
John Darrington [Tue, 28 Jul 2009 10:59:42 +0000 (12:59 +0200)]
remove unused return value

14 years agoAllow ranges with left to right and bottom to top sense to be selected.
John Darrington [Tue, 28 Jul 2009 10:55:23 +0000 (12:55 +0200)]
Allow ranges with left to right and bottom to top sense to be selected.

14 years agoReplace safe_strcmp with g_strcmp0.
John Darrington [Tue, 28 Jul 2009 05:05:42 +0000 (07:05 +0200)]
Replace safe_strcmp with g_strcmp0.

14 years agoFix clipboard export of non-ascii strings.
John Darrington [Tue, 28 Jul 2009 04:46:24 +0000 (06:46 +0200)]
Fix clipboard export of non-ascii strings.

Fixed problem pasting sheet data from psppire to
openoffice.org and gnumeric which contained non-ascii
characters.  Tested with openoffice.org 2.4 and
gnumeric 1.8.3

14 years agoFix bug in casemapping long strings.
John Darrington [Mon, 27 Jul 2009 17:41:25 +0000 (19:41 +0200)]
Fix bug in casemapping long strings.

Thanks to Ben Pfaff for pointing out the error.

14 years agoRemove the RESIZE feature of the sheet.
John Darrington [Mon, 27 Jul 2009 11:56:03 +0000 (13:56 +0200)]
Remove the RESIZE feature of the sheet.

This feature is currently unused, and probably
broken.

14 years agoRe-enable signal emmission on row/column selection
John Darrington [Mon, 27 Jul 2009 11:38:13 +0000 (13:38 +0200)]
Re-enable signal emmission on row/column selection

14 years agoRemoved unused function
John Darrington [Mon, 27 Jul 2009 10:58:06 +0000 (12:58 +0200)]
Removed unused function

14 years agoRemoved unnecessary variable and consequently unreachable code.
John Darrington [Mon, 27 Jul 2009 10:56:13 +0000 (12:56 +0200)]
Removed unnecessary variable and consequently unreachable code.