[PATCH 0/1] MR3577: ieframe/tests: Use less strict checks in session count tests.

[PATCH 1/1] ieframe/tests: Use less strict checks in session count tests.

[Jacek Caban]

From: Jacek Caban

---
 dlls/ieframe/tests/webbrowser.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/dlls/ieframe/tests/webbrowser.c b/dlls/ieframe/tests/webbrowser.c
index 15ce0272654..402014a8c07 100644
--- a/dlls/ieframe/tests/webbrowser.c
+++ b/dlls/ieframe/tests/webbrowser.c
@@ -4434,22 +4434,22 @@ static void test_SetAdvise(void)

static void test_SetQueryNetSessionCount(void)
 {
-    LONG count, init_count;
+    LONG count, init_count, inc_count, dec_count;

init_count = pSetQueryNetSessionCount(SESSION_QUERY);
     trace("init_count %ld\n", init_count);

-    count = pSetQueryNetSessionCount(SESSION_INCREMENT);
-    ok(count == init_count + 1, "count = %ld\n", count);
+    inc_count = pSetQueryNetSessionCount(SESSION_INCREMENT);
+    ok(inc_count > init_count, "count = %ld\n", inc_count);

count = pSetQueryNetSessionCount(SESSION_QUERY);
-    ok(count == init_count + 1, "count = %ld\n", count);
+    ok(count == inc_count, "count = %ld\n", count);

-    count = pSetQueryNetSessionCount(SESSION_DECREMENT);
-    ok(count == init_count, "count = %ld\n", count);
+    dec_count = pSetQueryNetSessionCount(SESSION_DECREMENT);
+    ok(dec_count < inc_count, "count = %ld\n", dec_count);

count = pSetQueryNetSessionCount(SESSION_QUERY);
-    ok(count == init_count, "count = %ld\n", count);
+    ok(count == dec_count, "count = %ld\n", count);
 }

static HRESULT WINAPI outer_QueryInterface(IUnknown *iface, REFIID riid, void **ppv)

[Francois Gouget (＠fgouget)]

Francois Gouget (@fgouget) commented about dlls/ieframe/tests/webbrowser.c:
> static void test_SetQueryNetSessionCount(void)
>  {
> - LONG count, init_count;
> -
>
> LONG count, init_count, inc_count, dec_count;
>
> init_count = pSetQueryNetSessionCount(SESSION_QUERY);
> trace("init_count %ld\n", init_count);
> - count = pSetQueryNetSessionCount(SESSION_INCREMENT);
> - ok(count == init_count + 1, "count = %ld\n", count);
> -
>
> inc_count = pSetQueryNetSessionCount(SESSION_INCREMENT);
> -
>
> ok(inc_count > init_count, "count = %ld\n", inc_count);
>
> count = pSetQueryNetSessionCount(SESSION_QUERY);
> - ok(count == init_count + 1, "count = %ld\n", count);
> - ok(count == inc_count, "count = %ld\n", count);

If IE creates its own net session between the SESSION_INCREMENT call and the SESSION_QUERY one this last ok() call will fail.

[Francois Gouget (＠fgouget)]

Francois Gouget (@fgouget) commented about dlls/ieframe/tests/webbrowser.c:
>  init_count = pSetQueryNetSessionCount(SESSION_QUERY);
>  trace("init_count %ld\n", init_count);
>
> - count = pSetQueryNetSessionCount(SESSION_INCREMENT);
> - ok(count == init_count + 1, "count = %ld\n", count);
> -
>
> inc_count = pSetQueryNetSessionCount(SESSION_INCREMENT);
> -
>
> ok(inc_count > init_count, "count = %ld\n", inc_count);
>
> count = pSetQueryNetSessionCount(SESSION_QUERY);
> - ok(count == init_count + 1, "count = %ld\n", count);
> - ok(count == inc_count, "count = %ld\n", count);
> - count = pSetQueryNetSessionCount(SESSION_DECREMENT);
> - ok(count == init_count, "count = %ld\n", count);
> - dec_count = pSetQueryNetSessionCount(SESSION_DECREMENT);
> - ok(dec_count < inc_count, "count = %ld\n", dec_count);

If IE creates its own net session between the SESSION_INCREMENT call and the SESSION_DECREMENT one, this ok() call will fail.

[Francois Gouget (＠fgouget)]

I'm sure there are many other race conditions that could happen that would make this test fail.

[Francois Gouget (＠fgouget)]

The commit message should reference bug 54866.

[Francois Gouget (＠fgouget)]

On Wed Aug 16 13:14:42 2023 +0000, Francois Gouget wrote:
> I'm sure there are many other race conditions that could happen that
> would make this test fail.

Well, I did say in bug 54866 that it looks like the extra net session is not created after the first SESSION_INCREMENT call so I guess this patch could work. But I as far as I can tell that's just how the timing of the race condition works and I don't think we should rely on that.