update iml to the 1.0.3 release + our patches

[williamstein]

spkg: http://boxen.math.washington.edu/home/jdemeyer/spkg/iml-1.0.3.p0.spkg (spkg diff)

Depends on #14699

CC: @jpflori

Component: packages: standard

Keywords: spkg upgrade iml sd40.5

Author: Jeroen Demeyer

Reviewer: Jean-Pierre Flori

Merged: sage-5.11.beta2

Issue created by migration from https://trac.sagemath.org/ticket/748

[williamstein]

comment:1

I've reverted the iml package that will be distributed with Sage to version 1.0.1.p6 (from sage-2.8.4)
until this gets resolved.

[williamstein]

comment:2

The 1.0.1.p8 version is in sage-2.8.5

[sagetrac-mabshoff]

comment:5

A couple commets:

• Arne merged the nullspace patch and a couple of the other fixes. Some of the other fixes no longer apply
• spkg-install does odd things, i.e fall back to non-ATLAS which is no longer needed
• The SPKG needs some general cleanup
• on OSX without creating 'repl' the build on the vanilla sources fails

Arne is putting together some 1.0.3 release which we can then test in Sage

Cheers,

Michael

[nexttime]

comment:8

I'm closing this as a duplicate of #9568.

[nexttime]

Changed keywords from none to spkg upgrade

[nexttime]

Reviewer: Leif Leonhardy

[sagetrac-drkirkby]

comment:10

Is it not more sensible for the duplicated ticket to be closed, rather than the original one? That's the way gcc handle it. If you report something that's a duplicate of an older ticket, the newer ticket gets closed - not the older one.

Dave

[kini]

comment:11

I guess it depends on which ticket has the most useful comments / patches and the most relevant people CC'd to it. At least #13027 has an SPKG attached. #9568 and this ticket certainly have more useful comments than #13027, though.

[kini]

comment:12

I have no objection to reopening this ticket, anyway. In any case, once someone actually solves the problem, all three will be closed.

[kini]

Changed reviewer from Leif Leonhardy to none

[kini]

Changed keywords from spkg upgrade to spkg upgrade iml

[kiwifb]

comment:13

I have a patch for repl problem in sage-on-gentoo but all the autotool chain needs to be regenerated after you apply it.
https://github.com/cschwan/sage-on-gentoo/blob/master/sci-libs/iml/files/iml-1.0.3-repl_removal.patch

[kini]

comment:14

Thanks, that will help. We cannot run autotools at build time, only at packaging time, since we don't have autotools as a dependency of Sage, nor do we package it in Sage. So rather than including your patch I guess I should apply it, run autotools, and then save that diff as a patch to apply at build time.

Before seeing your diff I had solved the repl problem like this:

diff --git a/repl/Makefile.am b/repl/Makefile.am
index 4cf001b..7be2e86 100644
--- a/repl/Makefile.am
+++ b/repl/Makefile.am
@@ -1,4 +1,4 @@
 noinst_LTLIBRARIES      = librepl.la
-librepl_la_SOURCES   =
+librepl_la_SOURCES   = dummy.c
 librepl_la_LIBADD    = @LTLIBOBJS@
 
diff --git a/repl/Makefile.in b/repl/Makefile.in
index 5c65bfa..ba963ec 100644
--- a/repl/Makefile.in
+++ b/repl/Makefile.in
@@ -51,7 +51,7 @@ CONFIG_HEADER = $(top_builddir)/config.h
 CONFIG_CLEAN_FILES =
 LTLIBRARIES = $(noinst_LTLIBRARIES)
 librepl_la_DEPENDENCIES = @LTLIBOBJS@
-am_librepl_la_OBJECTS =
+am_librepl_la_OBJECTS = dummy.lo
 librepl_la_OBJECTS = $(am_librepl_la_OBJECTS)
 DEFAULT_INCLUDES = -I. -I$(srcdir) -I$(top_builddir)
 depcomp = $(SHELL) $(top_srcdir)/config/depcomp
@@ -172,7 +172,7 @@ sharedstatedir = @sharedstatedir@
 sysconfdir = @sysconfdir@
 target_alias = @target_alias@
 noinst_LTLIBRARIES = librepl.la
-librepl_la_SOURCES = 
+librepl_la_SOURCES = dummy.c
 librepl_la_LIBADD = @LTLIBOBJS@
 all: all-am
 
@@ -226,6 +226,7 @@ distclean-compile:
 	-rm -f *.tab.c
 
 @AMDEP_TRUE@@am__include@ @am__quote@$(DEPDIR)/realloc.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/dummy.Plo@am__quote@
 
 .c.o:
 @am__fastdepCC_TRUE@	if $(COMPILE) -MT $@ -MD -MP -MF "$(DEPDIR)/$*.Tpo" -c -o $@ $<; \
@@ -367,7 +368,7 @@ clean-am: clean-generic clean-libtool clean-noinstLTLIBRARIES \
 	mostlyclean-am
 
 distclean: distclean-am
-	-rm -rf $(DEPDIR)
+	-rm -rf $(DEPDIR) ./$(DEPDIR)
 	-rm -f Makefile
 distclean-am: clean-am distclean-compile distclean-generic \
 	distclean-libtool distclean-tags
@@ -393,7 +394,7 @@ install-man:
 installcheck-am:
 
 maintainer-clean: maintainer-clean-am
-	-rm -rf $(DEPDIR)
+	-rm -rf $(DEPDIR) ./$(DEPDIR)
 	-rm -f Makefile
 maintainer-clean-am: distclean-am maintainer-clean-generic
 
diff --git a/repl/dummy.c b/repl/dummy.c
new file mode 100644
index 0000000..39e7442
--- /dev/null
+++ b/repl/dummy.c
@@ -0,0 +1 @@
+void dummy () { return; }
diff --git a/src/Makefile.in b/src/Makefile.in
index e9ca293..a1f5a81 100644
--- a/src/Makefile.in
+++ b/src/Makefile.in
@@ -223,6 +223,7 @@ libiml_la_CFLAGS = $(AM_CFLAGS)
 libiml_la_LIBADD = $(EXTERNLIB) \
 	 	   $(top_builddir)/repl/librepl.la
 
+libiml_la_LDFLAGS = -version-info 1:0:1
 all: all-am
 
 .SUFFIXES:

The first hunk is manually edited, as is the hunk adding the file repl/dummy.c; the rest are created by running automake in the src directory.

There are a couple of things that worry me, such as the -version-info 1:0:1 near the end of the diff. Does that mean that autotools are misunderstanding the version of IML as 1.0.1 instead of 1.0.3? Why does a similar line not already exist in that file? As far as I can tell, I made sure I was using the same versions of autotools that the authors of IML were - autoconf 2.59, automake 1.9.6, and libtool 1.5.22.

Anyway, I'll try again with your patch instead of this dummy function and see what's different. This time I'll also try the whole autotools suite by running autoreconf -i (as recommended to me by Jeroen) - is this what you recommend too, fbissey?

[jdemeyer]

comment:36

not yet doctested but at least the spkg and its testsuite work on Linux and OS X.

[nexttime]

comment:37

I still get

./configure: line 18624: -O2: command not found

[jdemeyer]

comment:38

Replying to @nexttime:

I still get

./configure: line 18624: -O2: command not found

Fixed in new spkg.

Doctests on boxen.math pass.

[jpflori]

comment:39

I already checked the spkg looks sane and installs fine.
Sage still pass its (not long) tests.

I just have to check the changes in the patches and I'll give it positive review.

[jpflori]

comment:40

Why merge tinyatlas.patch into sage1.patch?

On top of that I don't really get the use of the tinyatlas patch.
The function it defines (in a header file) is used nor in IML nor in Sage.
I'm testing Sage without it (no problems for IML).

So maybe renaming sage1.patch to memory.patch and discarding tinyatlas.patch would be enough.
And lets move the sage3_memleak.patch to memleak.patch while at it.

[jdemeyer]

comment:41

Replying to @jpflori:

Why merge tinyatlas.patch into sage1.patch?

Because those patches seem related: sage1.patch adds some include for tinyatlas.h.

On top of that I don't really get the use of the tinyatlas patch.
The function it defines (in a header file) is used nor in IML nor in Sage.

That's certainly not true. The files RNSop.c and certsolve.c in the src/src directory in IML use catlas_daxpby().

My guess is that's a function which normally comes from ATLAS, but which is needed on systems where ATLAS is not installed.

[jpflori]

comment:42

Replying to @jdemeyer:

Replying to @jpflori:

Why merge tinyatlas.patch into sage1.patch?

Because those patches seem related: sage1.patch adds some include for tinyatlas.h.

Ok, i thought the include was in tinyatlas.h as well, so I did not check.

On top of that I don't really get the use of the tinyatlas patch.
The function it defines (in a header file) is used nor in IML nor in Sage.

That's certainly not true. The files RNSop.c and certsolve.c in the src/src directory in IML use catlas_daxpby().

It seems to be true with IML 1.0.3 but not with 1.0.1 indeed.
(I mean the daxpby function was needed in 1.0.1 but is not anymore in 1.0.3.)

My guess is that's a function which normally comes from ATLAS, but which is needed on systems where ATLAS is not installed.

[jpflori]

Reviewer: Jean-Pierre Flori

[jpflori]

Work Issues: remove tinyatlas stuff, rename sage.patch*

[jdemeyer]

comment:44

Replying to @jpflori:

It seems to be true with IML 1.0.3 but not with 1.0.1.

OK, right, I checked the old sources.

[jdemeyer]

comment:45

I confirm that catlas_daxpby() is indeed nowhere used outside of the ATLAS sources.

Working on new spkg...

[jdemeyer]

comment:46

Applying the following changes:

diff --git a/SPKG.txt b/SPKG.txt
--- a/SPKG.txt
+++ b/SPKG.txt
@@ -32,21 +32,27 @@

 === Patches ===

- * rename_lift.patch: Change lift to iml_lift in padiclift.* and
-   nonsingsolve.*, since otherwise  on OSX you'll have horrible weird
-   conflict problems.
+ * blas_headers.patch: Add BLAS header files from GSL, needed in case
+   ATLAS has not been installed.
+ * build.patch: Made build scripts that work.
  * configure_default_cflags.patch: get rid of the following error
    during configure:
      ./configure: line 18624: -O3: command not found
- * Modified some of the examples, and made build scripts that work.
+ * examples.patch: Modified some of the examples.
+ * memleak.patch: use mpz_set_ui instead of mpz_init_set_ui on mpz
+   which is already allocated.
+ * remove_repl.patch: Do not build/install src/repl at all, since it
+   does nothing anyway and creating empty archives fails on OS X.

 == Changelog ==

 === iml-1.0.3.p0 (Jeroen Demeyer, 12 June 2013) ===
  * #748: Upgrade to latest upstream version, rebase patches.
  * Remove rename_lift.patch and sage2.patch, which were upstreamed.
- * Merged tinyatlas.patch into sage1.patch.
- * Apply sage3_memleak.patch in all 3 places with similar code.
+ * Removed tinyatlas.patch and #include "tinyatlas.h"
+ * Removed sage1.patch
+ * Apply sage3_memleak.patch in all 3 places with similar code, rename
+   to memleak.patch
  * Use -O3 optimization level by default.
  * Add configure_default_cflags.patch.

diff --git a/patches/sage3_memleak.patch b/patches/memleak.patch
rename from patches/sage3_memleak.patch
rename to patches/memleak.patch
diff --git a/patches/sage1.patch b/patches/sage1.patch
deleted file mode 100644
--- a/patches/sage1.patch
+++ /dev/null
@@ -1,80 +0,0 @@
-diff -ruN b/src/RNSop.c a/src/RNSop.c
---- b/src/RNSop.c      2006-11-23 22:25:23.000000000 +0100
-+++ a/src/RNSop.c      2013-06-10 23:05:18.872404179 +0200
-@@ -46,6 +46,7 @@
- 
- 
- #include "RNSop.h"
-+#include "tinyatlas.h"
- 
- /*
-  *
-diff -ruN b/src/memalloc.c a/src/memalloc.c
---- b/src/memalloc.c   2006-11-23 22:25:23.000000000 +0100
-+++ a/src/memalloc.c   2013-06-10 23:05:18.872404179 +0200
-@@ -48,13 +48,16 @@
- 
- #include "error.h"
- #include "common.h"
-+#include "stdio.h"
- 
- void *
- xmalloc (size_t num)
- {
-   void * new = malloc(num);
--  if (!new)
--    iml_fatal ("Memory exhausted");
-+  if (!new) {
-+    printf("%ul\n", num);
-+    iml_fatal ("Memory exhausted in xmalloc");
-+  }
-   return new;
- }
- 
-@@ -65,8 +68,10 @@
-   if (!p)
-     return xmalloc(num);
-   new = realloc(p, num);
--  if (!new)
--    iml_fatal("Memory exhausted");
-+  if (!new) {
-+    printf("%ul\n", num);
-+    iml_fatal("Memory exhausted in xrealloc");
-+  }
-   return new;
- }
- 
-@@ -76,8 +81,10 @@
- {
- #if HAVE_CALLOC
-   void * new = calloc(num, size);
--  if (!new)
--    iml_fatal("Memory exhausted");
-+  if (!new) {
-+    printf("%ul\n", num);
-+    iml_fatal("Memory exhausted in xcalloc");
-+  }
- #else
-   void * new = xmalloc(num*size);
-   bzero(new, num*size);
-diff -ruN iml-1.0.1-sage/src/tinyatlas.h src/src/tinyatlas.h
---- iml-1.0.1-sage/src/tinyatlas.h     1970-01-01 01:00:00.000000000 +0100
-+++ src/src/tinyatlas.h        2007-03-01 04:11:42.000000000 +0100
-@@ -0,0 +1,17 @@
-+/* 
-+Compute Y = alpha * X + beta * Y
-+
-+where 
-+   N = degree of each vector
-+   incX = X stride
-+   incY = Y stride
-+*/
-+
-+void catlas_daxpby(const int N, const double alpha, const double *X,
-+const int incX, const double beta, double *Y, const int incY) 
-+{
-+  int i;
-+  for(i=0; i < N; i++) {
-+    Y[i*incY] = alpha * X[i*incX] + beta * Y[i*incY];
-+  }
-+}

[jdemeyer]

Attachment: iml-1.0.3.p0.diff.gz

[jdemeyer]

Changed work issues from remove tinyatlas stuff, rename sage.patch* to none

[jdemeyer]

Merged: sage-5.11.beta2

[nexttime]

comment:50

Sign of life of IML upstream on #14648 !!1!111!

[nexttime]

comment:51

Replying to @nexttime:

Sign of life of IML upstream on #14648 !!1!111!

#14648 comment:51 ff. that is.