paweld/fpc
1
0
Fork 0
mirror of https://gitlab.com/freepascal.org/fpc/source.git synced 2025-04-07 07:28:26 +02:00
Code Issues Packages Projects Releases Wiki Activity

llvmdbg: fix overflows for aggregates > 2^61 bytes

Browse Source 
LLVM does not support aggregates larger than that at all, because internally
it stores all sizes in bits in an uint64. Their rationale for not having
special support for that is that there is no hardware with full 64 bit VM
address space anyway. So truncate our size emissions in debug info also to
that.
This commit is contained in:
Jonas Maebe 2022-05-29 13:22:46 +02:00
parent 9e7895575e
commit a19deace45
1 changed files with 14 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
compiler/llvm/dbgllvm.pas
View File

@ -945,7 +945,13 @@ implementation
if is_vector(def) then
dinode.addenum('flags','DIFlagVector');
if not is_dynamic_array(def) then
dinode.addqword('size',def.size*8)
if def.size<(qword(1) shl 61) then
dinode.addqword('size',def.size*8)
else
{ LLVM internally "only" supports sizes up to 1 shl 61, because they
store all sizes in bits in a qword; the rationale is that there
is no hardware supporting a full 64 bit address space either }
dinode.addqword('size',qword(1) shl 61)
else
begin
exprnode:=tai_llvmspecialisedmetadatanode.create(tspecialisedmetadatanodekind.DIExpression);
@ -974,7 +980,13 @@ implementation
dinode.addint64('tag',ord(DW_TAG_structure_type));
if (name<>'') then
dinode.addstring('name',name);
dinode.addqword('size',def.size*8);
if def.size<(qword(1) shl 61) then
dinode.addqword('size',def.size*8)
else
{ LLVM internally "only" supports sizes up to 1 shl 61, because they
store all sizes in bits in a qword; the rationale is that there
is no hardware supporting a full 64 bit address space either }
dinode.addqword('size',qword(1) shl 61);
list.concat(dinode);