[PATCH 1/2] server: Wait for process exit more often and using increasing delay.

[Rémi Bernon]

Instead of waiting for 1s before considering it dead, when most of the
time the process dies quickly, delaying desktop or wineserver shutdown.

This saves ~2s on the prefix shutdown time, as measured with:

time bash -c 'wine cmd /c exit |& tee /dev/null'

Signed-off-by: Rémi Bernon 
---
 server/process.c | 19 +++++++++++++++----
 server/process.h |  1 +
 2 files changed, 16 insertions(+), 4 deletions(-)

diff --git a/server/process.c b/server/process.c
index c8b32bd98b5..9e9d3f094cd 100644
--- a/server/process.c
+++ b/server/process.c
@@ -595,10 +595,20 @@ static void process_died( struct process *process )
 static void process_sigkill( void *private )
 {
     struct process *process = private;
+    int signal = 0;

-    process->sigkill_timeout = NULL;
-    kill( process->unix_pid, SIGKILL );
-    process_died( process );
+    process->sigkill_delay *= 2;
+    if (process->sigkill_delay >= TICKS_PER_SEC / 2)
+        signal = SIGKILL;
+
+    if (!kill( process->unix_pid, signal ) && !signal)
+        process->sigkill_timeout = add_timeout_user( -process->sigkill_delay, process_sigkill, process );
+    else
+    {
+        process->sigkill_delay = TICKS_PER_SEC / 64;
+        process->sigkill_timeout = NULL;
+        process_died( process );
+    }
 }

/* start the sigkill timer for a process upon exit */
@@ -606,7 +616,7 @@ static void start_sigkill_timer( struct process *process )
 {
     grab_object( process );
     if (process->unix_pid != -1)
-        process->sigkill_timeout = add_timeout_user( -TICKS_PER_SEC, process_sigkill, process );
+        process->sigkill_timeout = add_timeout_user( -process->sigkill_delay, process_sigkill, process );
     else
         process_died( process );
 }
@@ -630,6 +640,7 @@ struct process *create_process( int fd, struct process *parent, unsigned int fla
     process->handles         = NULL;
     process->msg_fd          = NULL;
     process->sigkill_timeout = NULL;
+    process->sigkill_delay   = TICKS_PER_SEC / 64;
     process->unix_pid        = -1;
     process->exit_code       = STILL_ACTIVE;
     process->running_threads = 0;
diff --git a/server/process.h b/server/process.h
index ef1c1206350..1cf554b920b 100644
--- a/server/process.h
+++ b/server/process.h
@@ -53,6 +53,7 @@ struct process
     process_id_t         group_id;        /* group id of the process */
     unsigned int         session_id;      /* session id */
     struct timeout_user *sigkill_timeout; /* timeout for final SIGKILL */
+    timeout_t            sigkill_delay;   /* delay before final SIGKILL */
     unsigned short       machine;         /* client machine type */
     int                  unix_pid;        /* Unix pid for final SIGKILL */
     int                  nice_limit;      /* RLIMIT_NICE of the process */

[PATCH 2/2] services: Terminate all service processes on shutdown.

[Rémi Bernon]

Instead of waiting for wineserver to kills them on shutdown, after a
2s delay.

This saves another ~2s on the prefix shutdown time, as measured with:

time bash -c 'wine cmd /c exit |& tee /dev/null'

Signed-off-by: Rémi Bernon 
---
 programs/services/services.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/programs/services/services.c b/programs/services/services.c
index 39cd9ee4ffa..d969cf81ab9 100644
--- a/programs/services/services.c
+++ b/programs/services/services.c
@@ -520,6 +520,7 @@ static void scmdatabase_wait_terminate(struct scmdatabase *db)
     {
         struct process_entry *process = grab_process(LIST_ENTRY(ptr, struct process_entry, entry));

+        process_terminate(process);
         scmdatabase_unlock(db);
         WaitForSingleObject(process->process, INFINITE);
         scmdatabase_lock(db);

[PATCH 2/2] services: Terminate all service processes on shutdown.

[Marvin]

Hi,

While running your changed tests, I think I found new failures.
Being a bot and all I'm not very good at pattern recognition, so I might be
wrong, but could you please double-check?

Full results can be found at:
https://testbot.winehq.org/JobDetails.pl?Key=101627

Your paranoid android.

=== debiant2 (build log) ===

error: patch failed: server/process.h:53
Task: Patch failed to apply

=== debiant2 (build log) ===

error: patch failed: server/process.h:53
Task: Patch failed to apply

[Marvin]

Hi,

While running your changed tests, I think I found new failures.
Being a bot and all I'm not very good at pattern recognition, so I might be
wrong, but could you please double-check?

Full results can be found at:
https://testbot.winehq.org/JobDetails.pl?Key=101626

Your paranoid android.

=== debiant2 (build log) ===

error: patch failed: server/process.h:53
Task: Patch failed to apply

=== debiant2 (build log) ===

error: patch failed: server/process.h:53
Task: Patch failed to apply