GNOME Bugzilla – Bug 723195
string.joinv() with empty array as argument goes really wrong
Last modified: 2014-03-03 09:54:26 UTC
After commit 3c0674fa (from bug 686451), the below code will allocate 18446744073709551613 bytes. That joinv is roughly what valadoc ends up doing when building rygel docs and a long comment happens to be empty: See bug 723102. public static int main (string[] args) { var s = string.joinv ("\n * ", "".split ("\n")); print ("%s\n", s); return ; }
Created attachment 267450 [details] [review] glib-2.0: Handle empty array as string.joinv() argument
(In reply to comment #1) > Created an attachment (id=267450) [details] [review] > glib-2.0: Handle empty array as string.joinv() argument Unfortunately that doesn't fix the issue for length == -1 (length will be != 0, so it will always get caught up in that condition instead of moving on to the == -1 check).
commit 0d8ab7be120106388b4c9777ad4715069a575228 Author: Evan Nemerson <evan@coeus-group.com> Date: Tue Jan 28 16:06:50 2014 -0800 glib-2.0: fix string.joinv for empty (but not null) arrays Fixes bug 723195.
(In reply to comment #2) > Unfortunately that doesn't fix the issue for length == -1 Oops. Thanks for a thorough and quick fix.
*** Bug 723102 has been marked as a duplicate of this bug. ***
Sorry for commenting on a closed bug, but it seems that commit 0d8ab7be introduced a new behavior when null is in the middle of the array, as in tests/methods/bug723195.vala: string[] a = { "foo", "bar", null, "baz" }; a.length = 4; assert (string.joinv (":", a) == "foo:bar::baz"); Is it really useful? As I wrote in bug 686451, previously I manually terminated the array with null to be sure. Now all such code results in a separator at the end.
You can set a.length = -1, that would make the array null terminated for string.joinv I guess.
Ah, yes. That would be the best workaround to support both newer and older valac. Thanks.