On Fri, Oct 11, 2019 at 04:22:57PM +0200, Stefan Sperling wrote:
> This happened during rebase and I haven't yet figured out why.
> In many instances it seems to work fine though, so I guess this
> is an expected result for some inputs.

Here is a real-world example where it worked great and helps to
understand what happened exactly.

In the code below, the branch being rebased refactored the function
iwm_enable_txq to iwm_enable_ac_txq, and changed IWM_CMD_QUEUE to
sc->cmdqid, and changed literal 7 to IWM_TX_FIFO_CMD.

The new base branch contained the latter change already which is
why diff3 flagged the conflict.


	iwm_write_prph(sc, IWM_SCD_CHAINEXT_EN, 0);

	/* enable command channel */
<<<<<<< commit dfff1dc915985d4cb44391fdcc2686cf098522cd
	err = iwm_enable_ac_txq(sc, sc->cmdqid, IWM_TX_FIFO_CMD);
	if (err)
		goto out;
|||||||
	err = iwm_enable_txq(sc, 0 /* unused */, IWM_CMD_QUEUE, 7);
	if (err)
		goto out;
=======
	err = iwm_enable_txq(sc, 0 /* unused */, IWM_CMD_QUEUE,
	    IWM_TX_FIFO_CMD);
	if (err) {
		iwm_nic_unlock(sc);
		return err;
	}
>>>>>>> sys/dev/pci/if_iwm.c

	/* Activate TX scheduler. */
	iwm_write_prph(sc, IWM_SCD_TXFACT, 0xff);