Bug 42532

Comment on attachment 53082 [details] [review]
Patch to build on Android

Review of attachment 53082 [details] [review]:
-----------------------------------------------------------------

Mostly looks fine, one query:

::: Android.mk
@@ +15,5 @@
> +	CPP=$(CONFIGURE_CPP) \
> +	CPPFLAGS="$(CONFIGURE_CPPFLAGS)" \
> +	PKG_CONFIG_LIBDIR="$(CONFIGURE_PKG_CONFIG_LIBDIR)" \
> +	PKG_CONFIG_TOP_BUILD_DIR=$(PKG_CONFIG_TOP_BUILD_DIR) \
> +	ac_cv_have_abstract_sockets=no \

Does Android really not have abstract sockets? They're a Linux kernel feature (originally from Solaris or something, I think), so even though Android doesn't have GNU userland, I'd expect abstract sockets to work.

(Abstract sockets are Unix sockets where the first byte of the address is \0 and all subsequent bytes are significant.)

(In reply to comment #1)
> > +	ac_cv_have_abstract_sockets=no \
> 
> Does Android really not have abstract sockets? They're a Linux kernel feature
> (originally from Solaris or something, I think), so even though Android doesn't
> have GNU userland, I'd expect abstract sockets to work.

Actually, never mind: this is dbus-glib, not dbus, so the abstract sockets check is completely pointless (the equivalent check in dbus is the one that matters). My branch in Bug #40711 discards it altogether; I'd appreciate review on that bug (preferably from a dbus reviewer, but anyone who knows Autotools really).

Fixed in git for 0.100.

(In reply to comment #2)
> Actually, never mind: this is dbus-glib, not dbus, so the abstract sockets
> check is completely pointless (the equivalent check in dbus is the one that
> matters). My branch in Bug #40711 discards it altogether; I'd appreciate review
> on that bug (preferably from a dbus reviewer, but anyone who knows Autotools
> really).

You know, even the one in dbus-1 is rather pointless because it tests the host system, not the target. The test should be done at runtime by trying to create an abstract socket and falling back to a regular one.