This fixes an issue where the debug locations for Swift traps were dropped in the produced PDB files, as they were pointing to line 0
I validated this on a sample project using WinDbgx, which can now correctly trap on the same line in multiple places
Many tests that use `-parse-stdlib` only do so because they want access
to the Builtin module. Now that we have the flag and ability to import
it, use that instead.
This patch replaces the stateful generation of SILScope information in
SILGenFunction with data derived from the ASTScope hierarchy, which should be
100% in sync with the scopes needed for local variables. The goal is to
eliminate the surprising effects that the stack of cleanup operations can have
on the current state of SILBuilder leading to a fully deterministic (in the
sense of: predictible by a human) association of SILDebugScopes with
SILInstructions. The patch also eliminates the need to many workarounds. There
are still some accomodations for several Sema transformation passes such as
ResultBuilders, which don't correctly update the source locations when moving
around nodes. If these were implemented as macros, this problem would disappear.
This necessary rewrite of the macro scope handling included in this patch also
adds proper support nested macro expansions.
This fixes
rdar://88274783
and either fixes or at least partially addresses the following:
rdar://89252827
rdar://105186946
rdar://105757810
rdar://105997826
rdar://105102288
The debugger relies on function arguments and local variables to be in different
scopes in order to disambiguate between local variables that shadow function
arguments.
rdar://83769198
Before this patch every Swift function would contain a top-level
DW_TAG_lexical_scope that didn't provide any useful information, used extra
space in the debug info and prevented local variables from showing up in virtual
async backtraces.
The SIL generation for this builtin also changes: instead of generating the cond_fail instructions upfront, let the optimizer generate it, if the operand is a static string literal.
In worst case, if the second operand is not a static string literal, the Builtin.condfail is lowered at the end of the optimization pipeline with a default message: "unknown program error".
The SIL generation for this builtin also changes: instead of generating the cond_fail instructions upfront, let the optimizer generate it, if the operand is a static string literal.
In worst case, if the second operand is not a static string literal, the Builtin.condfail is lowered at the end of the optimization pipeline with a default message: "unknown program error".
A workaround in codeview debuginfo generation was to declare a condfail
instruction as inlined to avoid using 0 as an artificial line location.
However, this was done without accounting for scopes that were already
legitimately inlined. So we'd end up with a condfail inlined from
function B into function A and then the IRGen for the condfail was
being given a debugloc claiming to have been inlined again. This was
causing a sitaution where we'd have debug info forfunctionA owning an
instruction which claimed it was owned by a different function.
Fix this by first checking if the `LastScope` was already inlined and,
if so, just used that scope instead.
