[PATCH 0/2] MR10072: shell32: Always update `outlen` (`pcchOut`) when querying strings

[Marijn Suijten (＠Marijn)]

At least `AssocQueryStringW()` is documented to update the character count in `pcchOut` to "the number of characters actually placed in the buffer" which requires us to update the value of `*outlen` even when the buffer size was sufficient: https://learn.microsoft.com/en-us/windows/win32/api/shlwapi/nf-shlwapi-assoc...

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=59402

-- 
https://gitlab.winehq.org/wine/wine/-/merge_requests/10072

[PATCH 1/2] shell32: Always update `outlen` (`pcchOut`) when querying strings

[Marijn Suijten]

From: Marijn Suijten <[e-mail]>

At least `AssocQueryStringW()` is documented to update the character
count in `pcchOut` to "the number of characters actually placed in the
buffer" which requires us to update the value of `*outlen` even when the
buffer size was sufficient:

https://learn.microsoft.com/en-us/windows/win32/api/shlwapi/nf-shlwapi-assoc...

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=59402
---
 dlls/shell32/assoc.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/dlls/shell32/assoc.c b/dlls/shell32/assoc.c
index 74df71241e4..fb96311eeef 100644
--- a/dlls/shell32/assoc.c
+++ b/dlls/shell32/assoc.c
@@ -464,7 +464,10 @@ static HRESULT ASSOC_ReturnString(ASSOCF flags, LPWSTR out, DWORD *outlen, LPCWS
         *outlen = datalen;
     }
     else
+    {
         len = datalen;
+        *outlen = datalen;
+    }
 
     if (len)
         memcpy(out, data, len*sizeof(WCHAR));
-- 
GitLab


https://gitlab.winehq.org/wine/wine/-/merge_requests/10072

[PATCH 2/2] shell32: Don't increase return-string length when truncating

[Marijn Suijten]

From: Marijn Suijten <[e-mail]>

According to
https://learn.microsoft.com/en-us/windows/win32/api/shlwapi/nf-shlwapi-assoc...
`pcchOut` should retain "the number of characters actually placed in
the buffer" which will be the maximum length of the buffer if `datalen`
(the actual length of the data to be returned) exceeds it. It seems that
behaviour shouldn't change unless `ASSOCF_NOTRUNCATE` is set (or `out`
is NULL) which doesn't return any data but only the actual required
length of the buffer.  Hence only update the `*pcchOut` argument
(represented by `*outlen` here) if the flag was set.

Also remove the unneeded `min()` as we already know that `*outlen` is
smaller than `datalen`.
---
 dlls/shell32/assoc.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/dlls/shell32/assoc.c b/dlls/shell32/assoc.c
index fb96311eeef..a9ee3dd6f4e 100644
--- a/dlls/shell32/assoc.c
+++ b/dlls/shell32/assoc.c
@@ -455,13 +455,13 @@ static HRESULT ASSOC_ReturnString(ASSOCF flags, LPWSTR out, DWORD *outlen, LPCWS
             len = 0;
             if (*outlen > 0) out[0] = 0;
             hr = E_POINTER;
+            *outlen = datalen;
         }
         else
         {
-            len = min(*outlen, datalen);
+            len = *outlen;
             hr = E_NOT_SUFFICIENT_BUFFER;
         }
-        *outlen = datalen;
     }
     else
     {
-- 
GitLab

https://gitlab.winehq.org/wine/wine/-/merge_requests/10072

[Marijn Suijten (＠Marijn)]

On Wed Apr 15 09:15:14 2026 +0000, Nikolay Sivov wrote:
> Please add some tests for this.
> Looking at https://gitlab.winehq.org/wine/wine/-/blob/master/dlls/shell32/tests/assoc.c, all tests rely on calling `GetString()` twice to acquire the length, rather than preallocating a likely-large-enough buffer and expecting `*pcchOut` to be updated to the returned length; will need to extend the tests for that.

It likely belongs in `test_IQueryAssociations_GetString()` although that only seems to test an initially-successful get for the length, and two failure cases with `E_NOT_SUFFICIENT_BUFFER` and `E_POINTER`.

-- 
https://gitlab.winehq.org/wine/wine/-/merge_requests/10072#note_136251