* gcj -C bug in xalan
@ 2001-09-09 9:41 Anthony Green
2001-09-10 13:24 ` Alexandre Petit-Bianco
` (2 more replies)
0 siblings, 3 replies; 8+ messages in thread
From: Anthony Green @ 2001-09-09 9:41 UTC (permalink / raw)
To: apbianco; +Cc: rhug-rhats
Hi Alex -
I recently checked in changes to build Xalan .class files before
the object code. For this to work, I had to add the following bits...
%.class: %.java
$(GCJ) -C --encoding=UTF-8 $(CLASSPATH) -O2 -d upstream/src $<
# Work around a gcj bug by putting the srcdir at the front of the classpath.
upstream/src/org/apache/xalan/xsltc/compiler/util/NodeType.class: upstream/src/org/apache/xalan/xsltc/compiler/util/NodeType.java
gcj -C --encoding=UTF-8 -fCLASSPATH=$(srcdir)/upstream/src:../xerces/upstream/src:../BCEL/upstream:../cup/upstream:$(srcdir)/upstream/src -O2 -d upstream/src $(srcdir)/upstream/src/org/apache/xalan/xsltc/compiler/util/NodeType.java
upstream/src/org/apache/xalan/xsltc/compiler/util/Type.class: upstream/src/org/apache/xalan/xsltc/compiler/util/Type.java
gcj -C --encoding=UTF-8 -fCLASSPATH=$(srcdir)/upstream/src:../xerces/upstream/src:../BCEL/upstream:../cup/upstream:$(srcdir)/upstream/src -O2 -d upstream/src $(srcdir)/upstream/src/org/apache/xalan/xsltc/compiler/util/Type.java
The default rule has the classpath starting with upstream/src (the
build directory). However, when building NoteType.class and
Type.class, gcj -C fails when it has to read generated class files
first. The exceptions above force the $(srcdir)/upstream/src to the
front of the classpath as a work around. If you remove them you'll
see errors like...
gcj -C --encoding=UTF-8 -fCLASSPATH=upstream/src:../xerces/upstream/src:../BCEL/upstream:../cup/upstream:/fencer/cutlass/aph/green/cvs/rhug/xalan/upstream/src -O2 -d upstream/src /fencer/cutlass/aph/green/cvs/rhug/xalan/upstream/src/org/apache/xalan/xsltc/compiler/util/Type.java
/fencer/cutlass/aph/green/cvs/rhug/xalan/upstream/src/org/apache/xalan/xsltc/compiler/NodeTest.java:79: Undefined variable or class name: `DOM'.
public static final int ANODE = DOM.FIRST_TYPE - 1;
^
/fencer/cutlass/aph/green/cvs/rhug/xalan/upstream/src/org/apache/xalan/xsltc/compiler/NodeTest.java:72: Undefined variable or class name: `DOM'.
public static final int ROOT = DOM.ROOT;
^
/fencer/cutlass/aph/green/cvs/rhug/xalan/upstream/src/org/apache/xalan/xsltc/compiler/NodeTest.java:73: Undefined variable or class name: `DOM'.
public static final int ELEMENT = DOM.ELEMENT;
^
/fencer/cutlass/aph/green/cvs/rhug/xalan/upstream/src/org/apache/xalan/xsltc/compiler/NodeTest.java:74: Undefined variable or class name: `DOM'.
public static final int ATTRIBUTE = DOM.ATTRIBUTE;
^
/fencer/cutlass/aph/green/cvs/rhug/xalan/upstream/src/org/apache/xalan/xsltc/compiler/NodeTest.java:69: Undefined variable or class name: `DOM'.
public static final int TEXT = DOM.TEXT;
^
/fencer/cutlass/aph/green/cvs/rhug/xalan/upstream/src/org/apache/xalan/xsltc/compiler/NodeTest.java:70: Undefined variable or class name: `DOM'.
public static final int COMMENT = DOM.COMMENT;
^
/fencer/cutlass/aph/green/cvs/rhug/xalan/upstream/src/org/apache/xalan/xsltc/compiler/NodeTest.java:71: Undefined variable or class name: `DOM'.
public static final int PI = DOM.PROCESSING_INSTRUCTION;
^
7 errors
make: *** [upstream/src/org/apache/xalan/xsltc/compiler/util/Type.class] Error 1
Any ideas?
AG
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: gcj -C bug in xalan
2001-09-09 9:41 gcj -C bug in xalan Anthony Green
@ 2001-09-10 13:24 ` Alexandre Petit-Bianco
2001-09-10 13:33 ` Anthony Green
2001-09-10 13:42 ` Anthony Green
2001-09-11 1:21 ` Alexandre Petit-Bianco
2001-09-13 15:14 ` Alexandre Petit-Bianco
2 siblings, 2 replies; 8+ messages in thread
From: Alexandre Petit-Bianco @ 2001-09-10 13:24 UTC (permalink / raw)
To: rhug-rhats
Anthony Green writes:
> Any ideas?
FYI, I'm looking into it. On the other hand, I'm getting some other
build errors:
cj -C --encoding=UTF-8 -fCLASSPATH=upstream/src:../xerces/upstream/src:../BCEL/upstream:../cup/upstream:/home/apbianco/src/rhug/xalan/upstream/src -O2 -d upstream/src /home/apbianco/src/rhug/xalan/upstream/src/javax/xml/transform/stream/StreamSource.java
gcj -C --encoding=UTF-8 -fCLASSPATH=upstream/src:../xerces/upstream/src:../BCEL/upstream:../cup/upstream:/home/apbianco/src/rhug/xalan/upstream/src -O2 -d upstream/src /home/apbianco/src/rhug/xalan/upstream/src/org/apache/xalan/extensions/ExpressionContext.java
/home/apbianco/src/rhug/xalan/upstream/src/org/apache/xml/utils/XMLString.java:648: unrecognized character in input stream.
* "Fahrvergn|gen".toUpperCase() returns "FAHRVERGN\GEN"
^
/home/apbianco/src/rhug/xalan/upstream/src/org/apache/xml/utils/XMLString.java:649: unrecognized character in input stream.
* "Visit Ljubinje!".toUpperCase() returns "VISIT LJUBINJE!"
^
/home/apbianco/src/rhug/xalan/upstream/src/org/apache/xpath/objects/XString.java:1044: unrecognized character in input stream.
* "Fahrvergn|gen".toUpperCase() returns "FAHRVERGN\GEN"
^
/home/apbianco/src/rhug/xalan/upstream/src/org/apache/xpath/objects/XString.java:1045: unrecognized character in input stream.
* "Visit Ljubinje!".toUpperCase() returns "VISIT LJUBINJE!"
^
4 errors
make[1]: *** [upstream/src/org/apache/xalan/extensions/ExpressionContext.class] Error 1
make[1]: Leaving directory `/raid/apbianco/build/rhug/i686-pc-linux-gnu/xalan'
make: *** [all-recursive] Error 1
I guess the trick would be to provide a different value for --encoding
(or no value at all) but then I get:
apbianco@janus[~/build/rhug/i686-pc-linux-gnu]: gcj -C -fCLASSPATH=upstream/src:../xerces/upstream/src:../BCEL/upstream:../cup/upstream:/home/apbianco/src/rhug/xalan/upstream/src -O2 -d upstream/src /home/apbianco/src/rhug/xalan/upstream/src/org/apache/xalan/extensions/ExpressionContext.java
jc1: No such file or directory: can't create directory upstream/src/org
./A
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: gcj -C bug in xalan
2001-09-10 13:24 ` Alexandre Petit-Bianco
@ 2001-09-10 13:33 ` Anthony Green
2001-09-10 13:42 ` Anthony Green
1 sibling, 0 replies; 8+ messages in thread
From: Anthony Green @ 2001-09-10 13:33 UTC (permalink / raw)
To: apbianco, rhug-rhats
Alex wrote:
> FYI, I'm looking into it. On the other hand, I'm getting some other
> build errors:
Interesting. I wonder why I didn't see those. Try changing --encoding=UTF-8
to --encoding=8859_1.
AG
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: gcj -C bug in xalan
2001-09-10 13:24 ` Alexandre Petit-Bianco
2001-09-10 13:33 ` Anthony Green
@ 2001-09-10 13:42 ` Anthony Green
1 sibling, 0 replies; 8+ messages in thread
From: Anthony Green @ 2001-09-10 13:42 UTC (permalink / raw)
To: apbianco, rhug-rhats
Alex wrote:
> FYI, I'm looking into it. On the other hand, I'm getting some other
> build errors:
I checked in a fix for this.
> apbianco@janus[~/build/rhug/i686-pc-linux-gnu]:
gcj -C -fCLASSPATH=upstream/src:../xerces/upstream/src:../BCEL/upstream:../cup
/upstream:/home/apbianco/src/rhug/xalan/upstream/src -O2 -d upstream/src
/home/apbianco/src/rhug/xalan/upstream/src/org/apache/xalan/extensions/Express
ionContext.java
> jc1: No such file or directory: can't create directory upstream/src/org
And this (in the toplevel configure.in/configure). For now, just create an
"xalan/upstream/src" directory in your build tree.
Thanks,
AG
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: gcj -C bug in xalan
2001-09-09 9:41 gcj -C bug in xalan Anthony Green
2001-09-10 13:24 ` Alexandre Petit-Bianco
@ 2001-09-11 1:21 ` Alexandre Petit-Bianco
2001-09-13 15:14 ` Alexandre Petit-Bianco
2 siblings, 0 replies; 8+ messages in thread
From: Alexandre Petit-Bianco @ 2001-09-11 1:21 UTC (permalink / raw)
To: rhug-rhats
Anthony Green writes:
> Any ideas?
Yeah, we're not keeping the information on imports when we leave the
scope of the class we're parsing. As a result, it shows when we try to
resolve type in some cases. I'm working on a patch.
./A
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: gcj -C bug in xalan
2001-09-09 9:41 gcj -C bug in xalan Anthony Green
2001-09-10 13:24 ` Alexandre Petit-Bianco
2001-09-11 1:21 ` Alexandre Petit-Bianco
@ 2001-09-13 15:14 ` Alexandre Petit-Bianco
2001-09-13 15:25 ` Anthony Green
2001-09-14 14:35 ` Anthony Green
2 siblings, 2 replies; 8+ messages in thread
From: Alexandre Petit-Bianco @ 2001-09-13 15:14 UTC (permalink / raw)
To: rhug-rhats
Anthony Green writes:
> Any ideas?
I appended a patch below (tested for regression and everything.) Also,
I found that it was necessary to revert this BCEL/Makefile.am patch to
let BCEL build:
Mon Sep 10 13:36:15 2001 Anthony Green <green@redhat.com>
* Makefile.am (GCJFLAGS): Remove "/src" from classpath.
* Makefile.in: Rebuilt.
With or without the compiler patch below. Any insights on why this
patch was needed in the first place?
Let me know about both patches.
./A
2001-09-13 Alexandre Petit-Bianco <apbianco@redhat.com>
* java-tree.h (TYPE_IMPORT_LIST): New macro.
(TYPE_IMPORT_DEMAND_LIST): Likewise.
(struct lang_type): New fields import_list and import_demand_list.
* parse.y (java_complete_class): Initialize TYPE_IMPORT_LIST and
TYPE_IMPORT_DEMAND_LIST with ctxp counterparts.
(do_resolve_class): New local saved_enclosing_type, initialized,
passed as parameter to find_in_imports and find_in_imports_on_demand.
(find_in_imports): Added paramater enclosing_type, use its
TYPE_IMPORT_LIST when applicable.
(find_in_imports_on_demand): Added parameter enclosing_type, use
its TYPE_IMPORT_DEMAND_LIST when applicable. Reorganized locals
declaration and initialization.
(fold_constant_for_init): Switch/restore current_class to the
appropriate context.
Index: java-tree.h
===================================================================
RCS file: /cvs/gcc/gcc/gcc/java/java-tree.h,v
retrieving revision 1.121
diff -u -p -r1.121 java-tree.h
--- java-tree.h 2001/09/07 18:40:26 1.121
+++ java-tree.h 2001/09/13 21:17:56
@@ -950,6 +950,8 @@ struct lang_decl_var
for non primitive types when compiling to bytecode. */
#define TYPE_DOT_CLASS(T) (TYPE_LANG_SPECIFIC(T)->dot_class)
#define TYPE_PACKAGE_LIST(T) (TYPE_LANG_SPECIFIC(T)->package_list)
+#define TYPE_IMPORT_LIST(T) (TYPE_LANG_SPECIFIC(T)->import_list)
+#define TYPE_IMPORT_DEMAND_LIST(T) (TYPE_LANG_SPECIFIC(T)->import_demand_list)
#define TYPE_PRIVATE_INNER_CLASS(T) (TYPE_LANG_SPECIFIC(T)->pic)
#define TYPE_PROTECTED_INNER_CLASS(T) (TYPE_LANG_SPECIFIC(T)->poic)
#define TYPE_HAS_FINAL_VARIABLE(T) (TYPE_LANG_SPECIFIC(T)->hfv)
@@ -968,6 +970,8 @@ struct lang_type
compiling to bytecode to implement
<non_primitive_type>.class */
tree package_list; /* List of package names, progressive */
+ tree import_list; /* Imported types, in the CU of this class */
+ tree import_demand_list; /* Imported types, in the CU of this class */
unsigned pic:1; /* Private Inner Class. */
unsigned poic:1; /* Protected Inner Class. */
unsigned hfv:1; /* Has final variables */
Index: parse.y
===================================================================
RCS file: /cvs/gcc/gcc/gcc/java/parse.y,v
retrieving revision 1.310
diff -u -p -r1.310 parse.y
--- parse.y 2001/09/13 14:37:24 1.310
+++ parse.y 2001/09/13 21:18:05
@@ -99,8 +99,8 @@ static tree parse_jdk1_1_error PARAMS ((
static void complete_class_report_errors PARAMS ((jdep *));
static int process_imports PARAMS ((void));
static void read_import_dir PARAMS ((tree));
-static int find_in_imports_on_demand PARAMS ((tree));
-static void find_in_imports PARAMS ((tree));
+static int find_in_imports_on_demand PARAMS ((tree, tree));
+static void find_in_imports PARAMS ((tree, tree));
static void check_static_final_variable_assignment_flag PARAMS ((tree));
static void reset_static_final_variable_assignment_flag PARAMS ((tree));
static void check_final_variable_local_assignment_flag PARAMS ((tree, tree));
@@ -5564,6 +5564,13 @@ java_complete_class ()
cclass = TREE_CHAIN (cclass), cclassd = CLASSD_CHAIN (cclassd))
{
jdep *dep;
+
+ /* We keep the compilation unit imports in the class so that
+ they can be used later to resolve type dependencies that
+ aren't necessary to solve now. */
+ TYPE_IMPORT_LIST (TREE_TYPE (cclass)) = ctxp->import_list;
+ TYPE_IMPORT_DEMAND_LIST (TREE_TYPE (cclass)) = ctxp->import_demand_list;
+
for (dep = CLASSD_FIRST (cclassd); dep; dep = JDEP_CHAIN (dep))
{
tree decl;
@@ -5750,6 +5757,7 @@ do_resolve_class (enclosing, class_type,
tree enclosing, class_type, decl, cl;
{
tree new_class_decl = NULL_TREE, super = NULL_TREE;
+ tree saved_enclosing_type = enclosing ? TREE_TYPE (enclosing) : NULL_TREE;
struct hash_table _ht, *circularity_hash = &_ht;
/* This hash table is used to register the classes we're going
@@ -5786,7 +5794,7 @@ do_resolve_class (enclosing, class_type,
/* 1- Check for the type in single imports. This will change
TYPE_NAME() if something relevant is found */
- find_in_imports (class_type);
+ find_in_imports (saved_enclosing_type, class_type);
/* 2- And check for the type in the current compilation unit */
if ((new_class_decl = IDENTIFIER_CLASS_VALUE (TYPE_NAME (class_type))))
@@ -5808,7 +5816,7 @@ do_resolve_class (enclosing, class_type,
/* 4- Check the import on demands. Don't allow bar.baz to be
imported from foo.* */
if (!QUALIFIED_P (TYPE_NAME (class_type)))
- if (find_in_imports_on_demand (class_type))
+ if (find_in_imports_on_demand (saved_enclosing_type, class_type))
return NULL_TREE;
/* If found in find_in_imports_on_demant, the type has already been
@@ -6710,17 +6718,22 @@ process_imports ()
statement. */
static void
-find_in_imports (class_type)
+find_in_imports (enclosing_type, class_type)
+ tree enclosing_type;
tree class_type;
{
- tree import;
-
- for (import = ctxp->import_list; import; import = TREE_CHAIN (import))
- if (TREE_VALUE (import) == TYPE_NAME (class_type))
- {
- TYPE_NAME (class_type) = EXPR_WFL_NODE (TREE_PURPOSE (import));
- QUALIFIED_P (TYPE_NAME (class_type)) = 1;
- }
+ tree import = (enclosing_type ? TYPE_IMPORT_LIST (enclosing_type) :
+ ctxp->import_list);
+ while (import)
+ {
+ if (TREE_VALUE (import) == TYPE_NAME (class_type))
+ {
+ TYPE_NAME (class_type) = EXPR_WFL_NODE (TREE_PURPOSE (import));
+ QUALIFIED_P (TYPE_NAME (class_type)) = 1;
+ return;
+ }
+ import = TREE_CHAIN (import);
+ }
}
static int
@@ -6869,14 +6882,17 @@ read_import_dir (wfl)
entire list, to detected potential double definitions. */
static int
-find_in_imports_on_demand (class_type)
+find_in_imports_on_demand (enclosing_type, class_type)
+ tree enclosing_type;
tree class_type;
{
- tree node, import, node_to_use = NULL_TREE;
+ tree import = (enclosing_type ? TYPE_IMPORT_DEMAND_LIST (enclosing_type) :
+ ctxp->import_demand_list);
+ tree node_to_use = NULL_TREE, cl = NULL_TREE;
+ tree node;
int seen_once = -1;
- tree cl = NULL_TREE;
- for (import = ctxp->import_demand_list; import; import = TREE_CHAIN (import))
+ while (import)
{
const char *id_name;
obstack_grow (&temporary_obstack,
@@ -6907,6 +6923,7 @@ find_in_imports_on_demand (class_type)
IDENTIFIER_POINTER (EXPR_WFL_NODE (TREE_PURPOSE (import))));
}
}
+ import = TREE_CHAIN (import);
}
if (seen_once == 1)
@@ -16129,8 +16146,14 @@ fold_constant_for_init (node, context)
}
else
{
+ /* Install the proper context for the field resolution.
+ The prior context is restored once the name is
+ properly qualified. */
+ tree saved_current_class = current_class;
/* Wait until the USE_COMPONENT_REF re-write. FIXME. */
+ current_class = DECL_CONTEXT (context);
qualify_ambiguous_name (node);
+ current_class = saved_current_class;
if (resolve_field_access (node, &decl, NULL)
&& decl != NULL_TREE)
return fold_constant_for_init (decl, decl);
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: gcj -C bug in xalan
2001-09-13 15:14 ` Alexandre Petit-Bianco
@ 2001-09-13 15:25 ` Anthony Green
2001-09-14 14:35 ` Anthony Green
1 sibling, 0 replies; 8+ messages in thread
From: Anthony Green @ 2001-09-13 15:25 UTC (permalink / raw)
To: apbianco, rhug-rhats
Alex wrote:
> With or without the compiler patch below. Any insights on why this
> patch was needed in the first place?
I have no idea :-( Coincidentally, I just noticed this also and committed
the fix.
Thanks -- I'll try your patch.
AG
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: gcj -C bug in xalan
2001-09-13 15:14 ` Alexandre Petit-Bianco
2001-09-13 15:25 ` Anthony Green
@ 2001-09-14 14:35 ` Anthony Green
1 sibling, 0 replies; 8+ messages in thread
From: Anthony Green @ 2001-09-14 14:35 UTC (permalink / raw)
To: apbianco, rhug-rhats
Alex wrote:
> Let me know about both patches.
This patch did the trick!
Thanks - I'll revert the work-around once you check in the gcj patch.
AG
^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2001-09-14 14:35 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2001-09-09 9:41 gcj -C bug in xalan Anthony Green
2001-09-10 13:24 ` Alexandre Petit-Bianco
2001-09-10 13:33 ` Anthony Green
2001-09-10 13:42 ` Anthony Green
2001-09-11 1:21 ` Alexandre Petit-Bianco
2001-09-13 15:14 ` Alexandre Petit-Bianco
2001-09-13 15:25 ` Anthony Green
2001-09-14 14:35 ` Anthony Green
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).