Allow to forward declare of mbedtls_blowfish_context #1215

[gelldur]

Status

READY

Requires Backporting

NO

Migrations

NO

[gelldur]

For issue: #1215

[gelldur]

Probably I should extend this PR for other such structs.

[yanesca]

@gelldur First of all thank you for your contribution! I agree, it would be best to extend other such structs too, because the issue (#1215) mentioned this as a general problem and blowfish was just an example.

[gelldur]

@yanesca I added (i think) all needed forward declares

[yanesca]

@gelldur Thanks for extending the fix! Could you please extend it a little bit further? I was thinking about these structures:

mbedtls_x509_crt_profile                                                        
mbedtls_havege_state                                                            
mbedtls_oid_descriptor_t                                                        
mbedtls_aes_xts_context                                                         
mbedtls_mpi                                                                     
mbedtls_pk_rsassa_pss_options                                                   
mbedtls_pk_debug_item                                                           
mbedtls_entropy_source_state                                                    
mbedtls_net_context

[yanesca]

(There are some other structs in _internal headers, but those shouldn't be used by the application anyway and therefore it is safe to leave them as they are. Preventing the forward declaration even has a function in this case as it further discourages the use of the internal headers.)

[gelldur]

@yanesca I updated. I'm not sure about your last comment. Did I add to many forward declare or it is ok ?

[yanesca]

@gelldur Thank you! You didn't add too many, I was just justifying why I didn't ask you to add even more.

[yanesca]

@gelldur Could you please rebase to have a single commit with all the changes and explain in the commit message why we are doing these changes?

[gelldur]

@yanesca done

[yanesca]

Thank you! It looks good to me.

[RonEld]

Was it tested on other platforms, such as Windows?

[RonEld]

I am approving this, however I would like to know how other compilers react to this change

[simonbutcher]

Whilst reviewing this, I was wondering if we need some style guidelines or similar to give struct names a different name from the typedef's (eg. the idiom of typedef struct _mystruct {...} mystruct;), however, I found we already appear to reuse the same name, as you can see in
asn1.h.

As such, I'm approving the change, but I see no need to backport.

[simonbutcher]

As an observation, this names all unnamed struct's apart from pk_internal.h and cipher_internal.h, which are internal and don't need this change.