Changeset 948307e for final-system/multilib

Timestamp:

Jul 29, 2006, 3:47:06 PM (18 years ago)

Author:

Ken Moffat <zarniwhoop@…>

Children:

b6fe857

Parents:

cf0287c

Message:

Finish the fix for libtool-32 test failure, and move the detail back into multilib/libtool.xml because multilib referencing an item in ppc64 and ppc64 referencing a different item in multilib causes recursion which prevents it rendering.

Location:

final-system/multilib

Files:

• libtool-n32.xml (modified) (1 diff)
• libtool.xml (modified) (2 diffs)

final-system/multilib/libtool-n32.xml

@@ -31 +31 @@
 
     <xi:include xmlns:xi="http://www.w3.org/2003/XInclude"
-    href="libtool.xml"
+    href="../common/libtool.xml"
     xpointer="xpointer(//*[@os='c'])"/>
 
     <xi:include xmlns:xi="http://www.w3.org/2003/XInclude"
-    href="libtool.xml"
+    href="../common/libtool.xml"
     xpointer="xpointer(//*[@os='d'])"/>
 
-    <xi:include xmlns:xi="http://www.w3.org/2003/XInclude"
-    href="libtool.xml"
-    xpointer="xpointer(//*[@os='e'])"/>
+    <para os="e">To test the results, identify the correct emulation, then issue:
+    <userinput>make LDEMULATION=[emulation] check</userinput>. The correct
+    emulation will be elf32btsmipn32 for a big-endian machine and elf32ltsmipn32
+    for a little-endian machine.</para>
 
     <xi:include xmlns:xi="http://www.w3.org/2003/XInclude"

final-system/multilib/libtool.xml

@@ -10 +10 @@
 
   <title>Libtool-&libtool-version; 32 Bit Libraries</title>
+
+<!-- with the introduction of an LDEMULATION  on make check, to fix a test
+ failure which has appeared for unknown reasons, this text is now specific
+ to mips -o32. -->
 
   <indexterm zone="ch-system-libtool-32">
@@ -37 +41 @@
     xpointer="xpointer(//*[@os='d'])"/>
 
-    <xi:include xmlns:xi="http://www.w3.org/2003/XInclude"
-    href="../common/libtool.xml"
-    xpointer="xpointer(//*[@os='e'])"/>
+    <para os="e">To test the results, first identify whether you are on a big- or
+    little-endian machine.  SGI machines are usually big-endian, Cobalt are
+    usually little-endian.  If in doubt you can <command>echo ${MACHTYPE} | grep 'el-'
+    </command>- this will match a little-endian machine.</para>
+
+    <para os="e2">To test on a little-endian machine issue:
+    <userinput>make LDEMULATION=elf32ltsmip check</userinput>.</para>
+
+    <para os="e1">To test on a big-endian machine issue:
+    <userinput>make LDEMULATION=elf32btsmip check</userinput>.</para>
+
+    <variablelist os="c">
+      <title>The meaning of the override on make check:</title>
+
+      <varlistentry>
+        <term><parameter>LDEMULATION=[emulation]</parameter></term>
+        <!-- <xi:include xmlns:xi="http://www.w3.org/2003/XInclude"
+        href="../ppc64/libtool.xml"
+        xpointer="xpointer(//*[@os='c2'])"/> -->
+        <listitem os="c2">
+          <para>Libtool tends to do the wrong thing when building for multilib,
+          at least on the non-default size(s) of architecture.  These causes of
+          these errors are not well understood and they can appear, or disappear,
+          as a result of apparently innocuous other changes in the build.  In
+          this version of the book, one of the tests (pdemo-make) fails to link
+          because it tries to link the 32-bit objects against 64-bit system
+          libraries. This option enables the test to succeed without impacting
+          the other tests (compare the common alternative fixes of
+          <literal>LD="gcc ${BUILD32}"</literal> which causes far fewer tests
+          to be executed, and configuring with
+          <literal>LDFLAGS='-L/lib -L/usr/lib'</literal> which in this case
+          causes other tests to fail.)</para>
+        </listitem>
+      </varlistentry>
+
+    </variablelist>
 
     <xi:include xmlns:xi="http://www.w3.org/2003/XInclude"