Condition always false in x509_get_subject_alt_name

[irwir]

Comparing with zero could be slightly more efficient, but also it is easier to understand.

https://github.com/ARMmbed/mbedtls/blob/beec142010cd9f3a967a9a3a7b407b51841b542e/library/x509_crt.c#L648
It might be guessed that the intention was to check end>=*p, but the values are pointers, and thus unsigned.
In pointer arithmetics anything less than 1 is nothing but 0.
Hence the condition means end == *p,

In any case, the loop header ensures that *p < end.

[k-stachowiak]

The check is in fact redundant, but I disagree with the explanation. end - *p is of type ptrdiff_t which is a signed type. It could be negative if *p went beyond end.
That will not happen, but not because of the pointer arithmetic. It won't happen because of the loop condition that you have pointed out.

[irwir]

@k-stachowiak
Thanks for looking at the issue.

end - *p is of type ptrdiff_t which is a signed type.

You are right here. Idea of treating pointers as unsigned remained from old times; now this is implementation defined.
By the way, VS compiler would use signed comparison in case of (end - *p) > 1, but unsigned if the condition is end > *p + 1

It could be negative if *p went beyond end.

Or if the difference is above maximum positive value for prtdiff_t type?
That is why it might be better to avoid expressions such as end - *p (quite a few of these are in asn1parse.c).