start using const on internal arguments #665
Conversation
Add const cast warnings. Add const when the arguments do not need to be modified.
|
This lays the ground work for using const in more places |
| @@ -212,7 +212,9 @@ static srtp_err_status_t srtp_aes_gcm_nss_context_init(void *cv, | |||
| return (srtp_err_status_cipher_fail); | |||
| } | |||
|
|
|||
| SECItem key_item = { siBuffer, (unsigned char *)key, c->key_size }; | |||
| /* explicitly cast away const of key */ | |||
| SECItem key_item = { siBuffer, (unsigned char *)(uintptr_t)key, | |||
There was a problem hiding this comment.
I think the compiler is whing too much if this is an issue.
| @@ -254,7 +254,9 @@ static srtp_err_status_t srtp_aes_icm_nss_context_init(void *cv, | |||
| return srtp_err_status_bad_param; | |||
| } | |||
|
|
|||
| SECItem keyItem = { siBuffer, (unsigned char *)key, c->key_size }; | |||
| /* explicitly cast away const of key */ | |||
| SECItem keyItem = { siBuffer, (unsigned char *)(uintptr_t)key, | |||
| @@ -187,7 +187,8 @@ static srtp_err_status_t srtp_hmac_init(void *statev, | |||
| return srtp_err_status_bad_param; | |||
| } | |||
|
|
|||
| SECItem key_item = { siBuffer, (unsigned char *)key, key_len }; | |||
| /* explicitly cast away const of key */ | |||
| SECItem key_item = { siBuffer, (unsigned char *)(uintptr_t)key, key_len }; | |||
| (srtp_hdr_xtnd_t *)((uint32_t *)hdr + uint32s_in_rtp_header + | ||
| hdr->cc); | ||
| const srtp_hdr_xtnd_t *xtn_hdr = | ||
| (const srtp_hdr_xtnd_t *)((const uint32_t *)hdr + |
There was a problem hiding this comment.
Pre-existing, but the double cast here is also a little surprising that it might be necessary. const additions good.
There was a problem hiding this comment.
This is only because of the usage of uint32s_in_rtp_header, so needs needs the first cast to do the right pointer arithmetic. I have have another change coming that provides some helper functions for the hdr ext stuff as these cast are ugly and murmurous.
| int profile_len = ntohs(xtn_hdr->length); | ||
| rtp_header_len += profile_len * 4; | ||
| /* profile length counts the number of 32-bit words */ | ||
| if (*pkt_octet_len < rtp_header_len) | ||
| if (pkt_octet_len < rtp_header_len) |
There was a problem hiding this comment.
Wow. I'm surprised the compiler did not complain about this one.
There was a problem hiding this comment.
this was due to api change from int * to int
| &mki_size); | ||
| session_keys = | ||
| srtp_get_session_keys(stream, (const uint8_t *)hdr, | ||
| (unsigned int)*pkt_octet_len, &mki_size); |
There was a problem hiding this comment.
While this cast might be necessary, why do we have a mix of signed and unsigned integer use? That might be a good thing to harmonize. I personally prefer unsigned integers for length values.
There was a problem hiding this comment.
This is something we have been looking at and yes we need to harmonize on this and other types, it is coming :)
For length I would even go as far as using size_t every where including in the public api in v3!
Add const cast warnings.
Add const when the arguments do not need to be modified.