[PATCH 0/1] MR8918: wineboot: Fix a memory leak in create_computer_name_keys.

[Alex Henrie (＠alexhenrie)]

If the canonical hostname was "localhost", the address info was not
freed.

Also, since we're touching the same line that defines the address info
hints, make the hints `static const` so that the compiler knows it's
safe to put them in read-only memory if it wants.

Fixes: ca5a6d07dc92ba631b178ec175e6b3fd5295e3d6

[PATCH 1/1] wineboot: Fix a memory leak in create_computer_name_keys.

[Alex Henrie]

From: Alex Henrie

If the canonical hostname was "localhost", the address info was not
freed.

Also, since we're touching the same line that defines the address info
hints, make the hints `static const` so that the compiler knows it's
safe to put them in read-only memory if it wants.

Fixes: ca5a6d07dc92ba631b178ec175e6b3fd5295e3d6
---
 programs/wineboot/wineboot.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/programs/wineboot/wineboot.c b/programs/wineboot/wineboot.c
index 8359e5d4c44..94ef024e32a 100644
--- a/programs/wineboot/wineboot.c
+++ b/programs/wineboot/wineboot.c
@@ -827,12 +827,12 @@ static void create_dynamic_registry_keys(void)
 /* create the ComputerName registry keys */
 static void create_computer_name_keys(void)
 {
-    struct addrinfo hints = {0}, *res;
+    static const struct addrinfo hints = { .ai_flags = AI_CANONNAME };
+    struct addrinfo *res = NULL;
     char *dot, buffer[256], *name = buffer;
     HKEY key, subkey;

if (gethostname( buffer, sizeof(buffer) )) return;
-    hints.ai_flags = AI_CANONNAME;
     if (!getaddrinfo( buffer, NULL, &hints, &res ) &&
         res->ai_canonname && strcasecmp(res->ai_canonname, "localhost") != 0)
         name = res->ai_canonname;
@@ -841,7 +841,7 @@ static void create_computer_name_keys(void)
     else dot = name + strlen(name);
     SetComputerNameExA( ComputerNamePhysicalDnsDomain, dot );
     SetComputerNameExA( ComputerNamePhysicalDnsHostname, name );
-    if (name != buffer) freeaddrinfo( res );
+    if (res) freeaddrinfo( res );

if (RegOpenKeyW( HKEY_LOCAL_MACHINE, L"System\CurrentControlSet\Control\ComputerName", &key ))
         return;

[Jinoh Kang (＠iamahuman)]

Jinoh Kang (@iamahuman) commented about programs/wineboot/wineboot.c:
> /* create the ComputerName registry keys */
>  static void create_computer_name_keys(void)
>  {
> - struct addrinfo hints = {0}, *res;
> -
>
> static const struct addrinfo hints = { .ai_flags = AI_CANONNAME };
> -
>
> struct addrinfo *res = NULL;
> char *dot, buffer[256], *name = buffer;
> HKEY key, subkey;
>
> if (gethostname( buffer, sizeof(buffer) )) return;
> - hints.ai_flags = AI_CANONNAME;
> if (!getaddrinfo( buffer, NULL, &hints, &res ) &&
>     res->ai_canonname && strcasecmp(res->ai_canonname, "localhost") != 0)

I think we should avoid assuming the function leaves the out variable untouched on failure. Also, let's take this opportunity to split the condition:

```suggestion:-0+0
    if (getaddrinfo( buffer, NULL, &hints, &res )) res = NULL;
    else if (strcasecmp(res->ai_canonname, "localhost") != 0)
```

(I removed the NULL check for ai_canonname, seems it's set to eqal nodename rather than set to NULL.)

[Jinoh Kang (＠iamahuman)]

Jinoh Kang (@iamahuman) commented about programs/wineboot/wineboot.c:
> /* create the ComputerName registry keys */
>  static void create_computer_name_keys(void)
>  {
> - struct addrinfo hints = {0}, *res;
> - static const struct addrinfo hints = { .ai_flags = AI_CANONNAME };
> - struct addrinfo *res = NULL;

```suggestion:-0+0
    struct addrinfo *res;
```

[Jinoh Kang (＠iamahuman)]

Looks good otherwise.

[Elizabeth Figura (＠zfigura)]

Probably would be better just to remove the freeaddrinfo()?

[Jinoh Kang (＠iamahuman)]

On Mon Sep  8 23:29:04 2025 +0000, Elizabeth Figura wrote:
> Probably would be better just to remove the freeaddrinfo()?

That would add noise to Valgrind and leak detectors output, right? I think those tools are pretty useful otherwise.

[Elizabeth Figura (＠zfigura)]

On Mon Sep  8 23:29:04 2025 +0000, Jinoh Kang wrote:
> That would add noise to Valgrind and leak detectors output, right? I
> think those tools are pretty useful otherwise.

The consistent position is that we don't care about cleaning up memory in short-lived programs, and as a corollary that being able to use them for leak detection is not worthwhile.