DHowett-MSFT added a comment.
While this fixes the assertion failure and the immediate issue of whether
packing _works_ inside delay-parsed templates in a PCH, it does reveal a
follow-on issue that I can't quite trace out.
c++
template <typename T>
void foo() {
#pragma pack(push, 1)
#pragma pack(show)
#pragma pack(pop)
}
results in...
unterminated '#pragma pack (push, ...)' at end of file
I've stepped through ASTReader and verified that we do get all three expected
pack pragma nodes
{PSK_Push, ..., 1}
{PSK_Show}
{PSK_Pop, ...}
suggesting that there's something later down the line that messes up the pack
stack.
================
Comment at: clang/test/PCH/delayed-template-with-pragma-pack.cpp:14
+int func() {
+#pragma pack(push, 1)
+ struct s { short a; T b; };
----------------
mikerice wrote:
> Your test should include testing of the slot string too, so we are sure all
> fields are serialized and restored correctly. Otherwise this looks reasonable
> to me.
So, I've identified an additional wrinkle...
Unlike `#pragma omp`, `#pragma pack` and `... align` don't show up in the AST
dump/reprint and I cannot verify their continued existence; this is true _even
without delayed template expansion or PCH generation._
alignment/packing seem to not be stored as Attrs, so they don't get printed via
attr print on the AST nodes or (prettily) via the tablegen-generated pretty
printers.
Fixing this is somewhat beyond me at the moment. :)
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D143410/new/
https://reviews.llvm.org/D143410
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
