Tom Lane wrote:
> Heikki Linnakangas <heikki@enterprisedb.com> writes:
>> I guess you guys already found a solution that works, but there's yet
>> another function, "BIO *BIO_new_mem_buf(void *data, int len)", that we
>> could use. We could open and read the file all by ourselves into memory,
>> then call BIO_new_mem_buf and pass that to PEM_read_X509. No need to
>> pass around file pointers, and we could handle any file I/O errors
>> ourselves. Presumably certificates are never very big, so reading it all
>> in memory shouldn't be a problem.
>
>> BIO_new_mem_buf was introduced in OpenSSL 0.9.7. What versions do we
>> support?
>
> This seems like a good idea. To judge from the release history at
> http://www.openssl.org/news/
> the OpenSSL boys stopped supporting 0.9.6 in 2004, so I figure we
> don't have to support it either. But 0.9.7 is still a live release
> branch, so it'd be good if we could play nice with it.
>
> http://www.openssl.org/docs/crypto/BIO_s_mem.html
Not sure it buys us much more than we already have. Sure, we can handle
file I/O. But we still can't handle errors in the parsing of the file.
I think that if we can open the file, then the chance that we fail to
read it is extremely small, compared to the chance that we get a parsing
error.
//Magnus
