Get rid of warnings in 64-bit about libssh2 HOT 7 CLOSED

I agree, no warnings would be ideal. But remember that we cannot change any types in the public API/ABI so where unsigned int is used today we cannot freely change it to size_t.

In the (1) case we need a typecast in that macro. The (2) thing is part of the protocol, isn't it?

Feel free to submit pull requests to fix warnings.

from libssh2.

remember that we cannot change any types in the public API/ABI so where unsigned int is used today we cannot freely change it to size_t.

I'm primarily interested in a practical solution to my current problem, which is that I see lots of compiler warnings, all of which are extremely useful to find 64-bit bugs in general. So I don't want to disable them, not even for a single translation unit.

On the other hand this means I cannot use the libssh2 function macros since they are not type-correct. One mitigation might be to add type casts to the macros.

This leaves the problem to the _ex functions only, but at least I can then type-cast on my own code, which is still wordy and ugly since I would prefer an API without the "len" parameters, which seem about CPU-perf optimization only.

The (2) thing is part of the protocol, isn't it?

Yes, maybe it's time to upgrade to a newer version of the spec?
https://filezilla-project.org/specs/draft-ietf-secsh-filexfer-13.txt

Just to clarify, I'm no expert in this area at all, basically I know SFTP and libssh2 for a week now :)

Feel free to submit pull requests to fix warnings.

I'm a C++ programmer, so I would probably fail to get the C style right. Still I created a patch adding the C-casts, but chances are it won't work since I use code-formatting in my projects.
I'l try to attach it...

from libssh2.

Sure we could implement support for a newer SFTP version. I don't think anyone will object to that. But that is a rather big undertaking that I'm not personally prepared to do anytime soon. We'll welcome help and patches of course!

from libssh2.

Would it be reasonable to define a new type, say libssh2_size_t, and use that instead? Then it could be a configuration option whether to stay ABI-compatible with the deployed library (via typedef unsigned int libssh2_size_t) or to use the system size_t.

The latter would be useful to those embedding libssh2 who don't have to worry about ABI-compatibility, and it would also allow building it with the system size_t locally just to run tests.

from libssh2.

Seems like a reasonable approach to me!

from libssh2.

I wrote that up as a separate issue (#29) since it wouldn't actually address these compiler warnings in many cases, but (I think) it would be a good general cleanup.

from libssh2.

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

from libssh2.