On 10/22/19 4:09 PM, Daniel P. Berrangé wrote:
On Tue, Oct 22, 2019 at 03:57:07PM +0200, Michal Privoznik wrote:
These functions can't fail really. Drop checking of their retval then.
These are declared G_GNUC_WARN_UNUSED_RESULT so this will surely result in a broken build when bisecting.
When I added the abort-on-OOM logic I intentionally left the G_GNUC_WARN_UNUSED_RESULT annotation in place, as I didn't want us to go through the step of dropping the return value checking, instead wanting to go straight to g_strdup_printf in one go to reduce the number of patches.
With this in mind, IMHO the first half of this series is redundant and each patch should be squashed into the corresponding patch in the second half of this series.
D'oh! I had it in one big patch (which compiled just fine) and only after that I've split it into smaller patches. You suggestion makes sense, I can do that. But I know Jano is looking at patches so I'll wait for his review. Michal