In the following pdf object /N has value 'elementname' encoded as UTF-16BE 19 0 obj << /K 18 0 R /S /div /Type /StructElem /A << /P [ << /N <feff0065006c0065006d0065006e0074006e0061006d0065> /V <feff004100410041> >> ] /O /UserProperties >> /Pg 15 0 R /P 17 0 R >> endobj Attribute::getName() returns 0xfe 0xff 0x00..0x65 but the caller doesn't know the size, so it uses only 0xfe 0xff 0x00. I propose these patches: 1) another constructor Attribute(const char *nameA, Object *valueA, int lenA) 2) a new method GooString *getUserPropertyName() to keep backwards compatibility (getName() is in the public interface) 3) a patch to parseUserProperty --- 0.26.4/poppler/StructElement.cc 2014-09-11 18:28:21.000000000 +0200 +++ p0.26.4/poppler/StructElement.cc 2014-10-06 09:13:31.013375479 +0200 @@ -690,6 +690,23 @@ valueA->copy(&value); } +Attribute::Attribute(const char *nameA, Object *valueA, int lenA): + type(UserProperty), + owner(UserProperties), + revision(0), + name(nameA,lenA), + value(), + hidden(gFalse), + formatted(NULL) +{ + assert(valueA); + valueA->copy(&value); +} + + + + + Attribute::Attribute(Type type, Object *valueA): type(type), owner(UserProperties), // TODO: Determine corresponding owner from Type @@ -785,13 +802,17 @@ return entry ? entry->type : Unknown; } + Attribute *Attribute::parseUserProperty(Dict *property) { Object obj, value; const char *name = NULL; + int len = 0 ; - if (property->lookup("N", &obj)->isString()) + if (property->lookup("N", &obj)->isString()){ name = obj.getString()->getCString(); + len = obj.getString()->getLength(); + } else if (obj.isName()) name = obj.getName(); else { @@ -807,7 +828,7 @@ return NULL; } - Attribute *attribute = new Attribute(name, &value); + Attribute *attribute = new Attribute(name, &value,len) ; value.free(); obj.free(); --- 0.26.4/poppler/StructElement.h 2014-09-11 18:28:20.000000000 +0200 +++ p0.26.4/poppler/StructElement.h 2014-10-06 09:13:53.469376410 +0200 @@ -76,6 +76,9 @@ // Creates an UserProperty attribute, with an arbitrary name and value. Attribute(const char *name, Object *value); + // Creates an UserProperty attribute, with an arbitrary name of lenght len and value. + Attribute(const char *name, Object *value, int len); + GBool isOk() const { return type != Unknown; } // Name, type and value can be set only on construction. @@ -87,6 +90,9 @@ static Object *getDefaultValue(Type type); const char *getName() const { return type == UserProperty ? name.getCString() : getTypeName(); } + GooString *getUserPropertyName() const { return type == UserProperty ? name.copy() : NULL; } + + // The revision is optional, and defaults to zero. Guint getRevision() const { return revision; }
There is no compatibility in internal poppler classes. I think the class should work on GooString. Can you please work on such a patch?
This seems to be ok --- StructElement.h.orig 2014-10-06 21:05:53.439147479 +0200 +++ StructElement.h 2014-10-07 07:51:25.412753215 +0200 @@ -76,6 +76,9 @@ // Creates an UserProperty attribute, with an arbitrary name and value. Attribute(const char *name, Object *value); + // Creates an UserProperty attribute, with an arbitrary name of length len and value. + Attribute(const char *name, Object *value, int len); + GBool isOk() const { return type != Unknown; } // Name, type and value can be set only on construction. @@ -86,7 +89,7 @@ Object *getValue() const { return &value; } static Object *getDefaultValue(Type type); - const char *getName() const { return type == UserProperty ? name.getCString() : getTypeName(); } + GooString *getName() const { return type == UserProperty ? name.copy() : new GooString(getTypeName()); } // The revision is optional, and defaults to zero. Guint getRevision() const { return revision; } --- StructElement.cc.orig 2014-10-06 21:05:47.551147234 +0200 +++ StructElement.cc 2014-10-06 09:13:31.000000000 +0200 @@ -690,6 +690,23 @@ valueA->copy(&value); } +Attribute::Attribute(const char *nameA, Object *valueA, int lenA): + type(UserProperty), + owner(UserProperties), + revision(0), + name(nameA,lenA), + value(), + hidden(gFalse), + formatted(NULL) +{ + assert(valueA); + valueA->copy(&value); +} + + + + + Attribute::Attribute(Type type, Object *valueA): type(type), owner(UserProperties), // TODO: Determine corresponding owner from Type @@ -785,13 +802,17 @@ return entry ? entry->type : Unknown; } + Attribute *Attribute::parseUserProperty(Dict *property) { Object obj, value; const char *name = NULL; + int len = 0 ; - if (property->lookup("N", &obj)->isString()) + if (property->lookup("N", &obj)->isString()){ name = obj.getString()->getCString(); + len = obj.getString()->getLength(); + } else if (obj.isName()) name = obj.getName(); else { @@ -807,7 +828,7 @@ return NULL; } - Attribute *attribute = new Attribute(name, &value); + Attribute *attribute = new Attribute(name, &value,len) ; value.free(); obj.free();
Can you attach a pdf where this is needed?
Created attachment 107516 [details] A pdf where the vlaue of /N is encoded as utf-16be
So are you using getName somewhere? because as far as i can see nothing in poppler is.
I've commited a patch inspired by yours but that fixes a few things.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.