Expand the public API slightly #35
Conversation
Fixes lifetime elision picking an overly short lifetime.
| @@ -48,7 +48,7 @@ impl<'a> X509Extension<'a> { | |||
| } | |||
|
|
|||
| /// Return the extension type or `None` if the extension is not implemented. | |||
| pub fn parsed_extension(&self) -> &ParsedExtension { | |||
| pub fn parsed_extension(&self) -> &ParsedExtension<'a> { | |||
There was a problem hiding this comment.
I'd like to understand - the compiler was inferring a wrong lifetime without the explicit one?
There was a problem hiding this comment.
Indeed! Lifetime elision was picking up the anonymous lifetime on the &self reference, and not the 'a lifetime on the impl block.
My usage goes like this:
I have a long-lived ('a) object that carries the entire DER of the certificate.
A function parses it and returns &str data that is sliced from the same DER.
The &str should be able to live as long as the parent DER, but the X509Extension and ParsedExtension structs are local to the function; having the reference lifetime (function local) distinct from the 'a lifetime used inside the structs allows me to return the right &'a str.
There was a problem hiding this comment.
Ah, I see now, thank you for the details.
I also encountered this kind of limitations within DER, where the compiler did not apply "transitive" lifetimes to inner objects when returned, but agrees when I give it explicitly - maybe I will be able to fix this now!
|
|
||
| impl<'a> X509Name<'a> { | ||
| // Not using the AsRef trait, as that would not give back the full 'a lifetime | ||
| pub fn as_raw(&self) -> &'a [u8] { |
There was a problem hiding this comment.
- I'm also curious about the comment on lifetimes (which seems similar to the above one)
- Could you share the use case where you needed the raw bytes? I'm asking mostly because, if this is useful for several fields, the
as_rawmethod could be added to all parsed components
There was a problem hiding this comment.
Similar to the above; the AsRef trait takes no lifetime parameter, and the as_ref function can only return data with the lifetime on the X509Name; if the parsing is done in an inner function, the inner function won't be able to return slices from that.
My use case is sending OCSP requests and validating the reponses; the raw data is sometimes hashed (building the CertID for the OCSP protocol), it can also be copied (some parts of the request should exactly match the certificates that are being checked), and compared (to check that the response really matches the request).
| @@ -51,14 +51,23 @@ fn parse_rdn(i: &[u8]) -> BerResult<RelativeDistinguishedName> { | |||
| .map(|(rem, x)| (rem, x.1)) | |||
| } | |||
|
|
|||
| pub(crate) fn parse_name(i: &[u8]) -> BerResult<X509Name> { | |||
| pub fn parse_name(i: &[u8]) -> BerResult<X509Name> { | |||
There was a problem hiding this comment.
I agree with making this public, but FYI I think I'll just change to a better name (parse_x509_name) by adding a commit
|
I think having as_raw everywhere would prove useful. |
I'll see if this can be added easily, but that will be a different issue. Just thinking of rewriting |
|
Merged, thanks! |
For X509Name, exposes parsing as well as accessing the raw data after parsing.
For parsed certificate extensions, this fixes an unnecessarily short lifetime.