| Summary: | crash because of null string on solaris | ||
|---|---|---|---|
| Product: | PulseAudio | Reporter: | Pierre Ossman <pierre-bugzilla> |
| Component: | core | Assignee: | pulseaudio-bugs |
| Status: | RESOLVED FIXED | QA Contact: | pulseaudio-bugs |
| Severity: | normal | ||
| Priority: | medium | CC: | astrand, lennart |
| Version: | unspecified | ||
| Hardware: | Other | ||
| OS: | All | ||
| Whiteboard: | |||
| i915 platform: | i915 features: | ||
If there's anything we can do to make it easier to integrate this work, please let us know. Would a git pull request help? Sorry for ignoring this. There was shortage of maintainer time prior to last November. This particular change seems strange to me. "git grep pa_thread_new" indicates that threads are always created with a name, so t->name should never be NULL. As for handling possibly-NULL strings in printf arguments in general, we have pa_strnull(), which may be in many cases more appropriate than patching string-returning functions to never return NULL. Threads not created by PulseAudio however can lack a name. In this case the main thread.
Alternative patch using pa_strnull():
Index: src/pulsecore/log.c
===================================================================
--- src/pulsecore/log.c (revision 27649)
+++ src/pulsecore/log.c (working copy)
@@ -303,9 +303,9 @@
pa_vsnprintf(text, sizeof(text), format, ap);
if ((_flags & PA_LOG_PRINT_META) && file && line > 0 && func)
- pa_snprintf(location, sizeof(location), "[%s][%s:%i %s()] ", pa_thread_get_name(pa_thread_self()), file, line, func);
+ pa_snprintf(location, sizeof(location), "[%s][%s:%i %s()] ", pa_strnull(pa_thread_get_name(pa_thread_self())), file, line, func);
else if ((_flags & (PA_LOG_PRINT_META|PA_LOG_PRINT_FILE)) && file)
- pa_snprintf(location, sizeof(location), "[%s] %s: ", pa_thread_get_name(pa_thread_self()), pa_path_get_filename(file));
+ pa_snprintf(location, sizeof(location), "[%s] %s: ", pa_strnull(pa_thread_get_name(pa_thread_self())), pa_path_get_filename(file));
else
location[0] = 0;
|
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.
Solaris' printf doesn't support NULL arguments for %s, which pulseaudio unfortunately gives it now and then. Making a safe printf wrapper is just a major pain, so I guess we'll have to find the offenders that provide null pointers. The most glaring one is pa_thread_get_name(): Index: src/pulsecore/thread-posix.c =================================================================== --- src/pulsecore/thread-posix.c (revision 25954) +++ src/pulsecore/thread-posix.c (working copy) @@ -211,6 +211,9 @@ } #endif + if (t->name == NULL) + return "(null)"; + return t->name; }