Author: Armin Rigo <[email protected]>
Branch: null_byte_after_str
Changeset: r85920:22471be160be
Date: 2016-07-30 10:34 +0200
http://bitbucket.org/pypy/pypy/changeset/22471be160be/
Log: Some fixes to STRGETITEM/STRSETITEM
diff --git a/rpython/jit/backend/llsupport/descr.py
b/rpython/jit/backend/llsupport/descr.py
--- a/rpython/jit/backend/llsupport/descr.py
+++ b/rpython/jit/backend/llsupport/descr.py
@@ -280,7 +280,7 @@
concrete_type = '\x00'
def __init__(self, basesize, itemsize, lendescr, flag, is_pure=False,
concrete_type='\x00'):
- self.basesize = basesize
+ self.basesize = basesize # this includes +1 for STR
self.itemsize = itemsize
self.lendescr = lendescr # or None, if no length
self.flag = flag
@@ -676,7 +676,7 @@
def unpack_arraydescr(arraydescr):
assert isinstance(arraydescr, ArrayDescr)
- ofs = arraydescr.basesize
+ ofs = arraydescr.basesize # this includes +1 for STR
size = arraydescr.itemsize
sign = arraydescr.is_item_signed()
return size, ofs, sign
diff --git a/rpython/jit/backend/llsupport/rewrite.py
b/rpython/jit/backend/llsupport/rewrite.py
--- a/rpython/jit/backend/llsupport/rewrite.py
+++ b/rpython/jit/backend/llsupport/rewrite.py
@@ -293,6 +293,7 @@
basesize, itemsize, ofs_length = get_array_token(rstr.STR,
self.cpu.translate_support_code)
assert itemsize == 1
+ basesize -= 1 # for the extra null character
self.emit_gc_load_or_indexed(op, op.getarg(0), op.getarg(1),
itemsize, itemsize, basesize,
NOT_SIGNED)
elif opnum == rop.UNICODEGETITEM:
@@ -304,6 +305,7 @@
basesize, itemsize, ofs_length = get_array_token(rstr.STR,
self.cpu.translate_support_code)
assert itemsize == 1
+ basesize -= 1 # for the extra null character
self.emit_gc_store_or_indexed(op, op.getarg(0), op.getarg(1),
op.getarg(2),
itemsize, itemsize, basesize)
elif opnum == rop.UNICODESETITEM:
diff --git a/rpython/jit/backend/llsupport/symbolic.py
b/rpython/jit/backend/llsupport/symbolic.py
--- a/rpython/jit/backend/llsupport/symbolic.py
+++ b/rpython/jit/backend/llsupport/symbolic.py
@@ -29,7 +29,7 @@
def get_array_token(T, translate_support_code):
# T can be an array or a var-sized structure
if translate_support_code:
- basesize = llmemory.sizeof(T, 0)
+ basesize = llmemory.sizeof(T, 0) # this includes +1 for STR
if isinstance(T, lltype.Struct):
SUBARRAY = getattr(T, T._arrayfld)
itemsize = llmemory.sizeof(SUBARRAY.OF)
@@ -57,6 +57,7 @@
assert carray.length.size == WORD
ofs_length = before_array_part + carray.length.offset
basesize = before_array_part + carray.items.offset
+ basesize += T._hints.get('extra_item_after_alloc', 0) # +1 for STR
carrayitem = ll2ctypes.get_ctypes_type(T.OF)
itemsize = ctypes.sizeof(carrayitem)
return basesize, itemsize, ofs_length
diff --git a/rpython/jit/backend/llsupport/test/test_rewrite.py
b/rpython/jit/backend/llsupport/test/test_rewrite.py
--- a/rpython/jit/backend/llsupport/test/test_rewrite.py
+++ b/rpython/jit/backend/llsupport/test/test_rewrite.py
@@ -647,6 +647,9 @@
""")
def test_rewrite_assembler_newstr_newunicode(self):
+ # note: strdescr.basesize already contains the extra final character,
+ # so that's why newstr(14) is rounded up to 'basesize+15' and not
+ # 'basesize+16'.
self.check_rewrite("""
[i2]
p0 = newstr(14)
@@ -657,12 +660,12 @@
""", """
[i2]
p0 = call_malloc_nursery( \
- %(strdescr.basesize + 16 * strdescr.itemsize + \
+ %(strdescr.basesize + 15 * strdescr.itemsize + \
unicodedescr.basesize + 10 * unicodedescr.itemsize)d)
gc_store(p0, 0, %(strdescr.tid)d, %(tiddescr.field_size)s)
gc_store(p0, %(strlendescr.offset)s, 14,
%(strlendescr.field_size)s)
gc_store(p0, 0, 0, %(strhashdescr.field_size)s)
- p1 = nursery_ptr_increment(p0, %(strdescr.basesize + 16 *
strdescr.itemsize)d)
+ p1 = nursery_ptr_increment(p0, %(strdescr.basesize + 15 *
strdescr.itemsize)d)
gc_store(p1, 0, %(unicodedescr.tid)d, %(tiddescr.field_size)s)
gc_store(p1, %(unicodelendescr.offset)s, 10,
%(unicodelendescr.field_size)s)
gc_store(p1, 0, 0, %(unicodehashdescr.field_size)s)
@@ -1240,14 +1243,14 @@
# 'i3 = gc_load_i(p0,i5,%(unicodedescr.itemsize)d)'],
[True, (4,), 'i3 = strgetitem(p0,i1)' '->'
'i3 = gc_load_indexed_i(p0,i1,1,'
- '%(strdescr.basesize)d,1)'],
+ '%(strdescr.basesize-1)d,1)'],
#[False, (4,), 'i3 = strgetitem(p0,i1)' '->'
- # 'i5 = int_add(i1, %(strdescr.basesize)d);'
+ # 'i5 = int_add(i1, %(strdescr.basesize-1)d);'
# 'i3 = gc_load_i(p0,i5,1)'],
## setitem str/unicode
[True, (4,), 'i3 = strsetitem(p0,i1,0)' '->'
'i3 = gc_store_indexed(p0,i1,0,1,'
- '%(strdescr.basesize)d,1)'],
+ '%(strdescr.basesize-1)d,1)'],
[True, (2,4), 'i3 = unicodesetitem(p0,i1,0)' '->'
'i3 = gc_store_indexed(p0,i1,0,'
'%(unicodedescr.itemsize)d,'
_______________________________________________
pypy-commit mailing list
[email protected]
https://mail.python.org/mailman/listinfo/pypy-commit
