On Mon, 2019-06-17 at 17:42 +0300, Vitaly Chikunov wrote: > > I like simple stories, where everything is laid out. Think of a patch > > set as a simple story, not a mystery, explaining how you go from point > > A to point B. Each patch in the series explains the next step. > > > > The changes in cmd_import() are nice, straight forward, and easy to > > read. Other than the "calc_keyid_v2()" change from passing an RSA key > > to pkey key" dependency, this could be a separate patch. > > Please see if you could further break this patch up to simplify review. > > I thought about splitting as much as possible, but didn't not find what > to split more. > > - If we add read_pub_pkey into cmd_import then we can not use old > calc_keyid_v2 (which wants RSA). > - If we add new calc_keyid_v2 into cmd_import it would require calling > read_pub_pkey (and have sign_hash_v2 updated too). > - We can not put cmd_import change before 'rsa to pkey' change, because > calc_keyid_v2 is also used in new sign_hash_v2. > - We can not put cmd_import change after 'rsa to pkey' change, because > cmd_import will fail for v2 signatures (which should use PKEY). > > So I don't see how this is splittable. Perhaps it causes more patch churn, but there always is a way. In this case, there are a couple of ways. Instead of replacing "key" with "pkey": - define pkey as an additional argument to calc_keyid_v2(). - define a new function named, for example, calc_keyid_v2_pkey(). Mimi
