[msi] Is there a potential problem with squash_guid in msi/registry.c ?

[Paul Vriens]

Hi,

I'm still looking into the fact why the msi.c tests fails for me (and succeeds 
when running with WINEDEBUG=+msi).

A trace (+msi,+reg,+ole) revealed:

trace:msi:MsiQueryProductStateW L"A938G02JF-2NF3N93-VN3-2NNF-3KGKALDNF93"
trace:msi:MSIREG_OpenUserProductsKey L"A938G02JF-2NF3N93-VN3-2NNF-3KGKALDNF93"
trace:reg:NtOpenKey 
(0x48,L"A938G02JF-2NF3N93-VN3-2NNF-3KGKALDNF93\\CLSID",f003f,0x34f37c)
trace:reg:NtOpenKey <- (nil)
warn:ole:CLSIDFromProgID couldn't open key for ProgID 
L"A938G02JF-2NF3N93-VN3-2NNF-3KGKALDNF93"
trace:msi:squash_guid Returning FALSE
trace:msi:MSIREG_OpenUserProductsKey squished 
(L"\7bc9\0480\0011\f8c84\2f47\7bc3\c7ea\7ecd\f9204\7139\7ecf\c89a\7ecd\00ff")

(the 'Returning FALSE' is my doing).

The squished_pc variable in MSIREG_OpenUserProductsKey is filled with garbage.

I think that we should patch msi/registry.c:

diff --git a/dlls/msi/registry.c b/dlls/msi/registry.c
index b9c0af4..24a3def 100644
--- a/dlls/msi/registry.c
+++ b/dlls/msi/registry.c
@@ -233,6 +233,8 @@ BOOL squash_guid(LPCWSTR in, LPWSTR out)
      DWORD i,n=1;
      GUID guid;

+    out[0] = 0;
+
      if (FAILED(CLSIDFromString((LPOLESTR)in, &guid)))
          return FALSE;

So 'out' is always clean at the start. Adding this patch makes the test fail 
consistently :-). So it's not the fix.

Any comments/remarks/suggestions ?

Cheers,

Paul.

[Paul Vriens]

Paul Vriens wrote:
> Hi,
>
> I'm still looking into the fact why the msi.c tests fails for me (and 
> succeeds when running with WINEDEBUG=+msi).
>
> A trace (+msi,+reg,+ole) revealed:
>
> trace:msi:MsiQueryProductStateW L"A938G02JF-2NF3N93-VN3-2NNF-3KGKALDNF93"
> trace:msi:MSIREG_OpenUserProductsKey 
> L"A938G02JF-2NF3N93-VN3-2NNF-3KGKALDNF93"
> trace:reg:NtOpenKey 
> (0x48,L"A938G02JF-2NF3N93-VN3-2NNF-3KGKALDNF93\\CLSID",f003f,0x34f37c)
> trace:reg:NtOpenKey <- (nil)
> warn:ole:CLSIDFromProgID couldn't open key for ProgID 
> L"A938G02JF-2NF3N93-VN3-2NNF-3KGKALDNF93"
> trace:msi:squash_guid Returning FALSE
> trace:msi:MSIREG_OpenUserProductsKey squished 
> (L"\7bc9\0480\0011\f8c84\2f47\7bc3\c7ea\7ecd\f9204\7139\7ecf\c89a\7ecd\00ff")
>
> (the 'Returning FALSE' is my doing).
>
> The squished_pc variable in MSIREG_OpenUserProductsKey is filled with 
> garbage.
>
> I think that we should patch msi/registry.c:
>
> diff --git a/dlls/msi/registry.c b/dlls/msi/registry.c
> index b9c0af4..24a3def 100644
> --- a/dlls/msi/registry.c
> +++ b/dlls/msi/registry.c
> @@ -233,6 +233,8 @@ BOOL squash_guid(LPCWSTR in, LPWSTR out)
>      DWORD i,n=1;
>      GUID guid;
>
> +    out[0] = 0;
> +
>      if (FAILED(CLSIDFromString((LPOLESTR)in, &guid)))
>          return FALSE;
>
> So 'out' is always clean at the start. Adding this patch makes the test 
> fail consistently :-). So it's not the fix.
>
> Any comments/remarks/suggestions ?
>
> Cheers,
>
> Paul.

Another question about squash_guid. Why does it return a BOOL but that 
information is not used by any of the callers?

Should we both patch squash_guid (to at least no return garbage) and change all 
the callers to bail out if squash_guid failed?

Cheers,

Paul.