[illumos-Developer] webrev #687: rpcgen should not generate absolute #includes

[Stepan Zastupov]

On Thu, May 12, 2011 at 7:11 PM, Gordon Ross <gordon.w.ross at gmail.com> wrote:
> On Thu, May 12, 2011 at 7:37 AM, Stepan Zastupov
> <stepan.zastupov at gmail.com> wrote:
>> Here is the updated patch:
>> http://cr.illumos.org/view/0ig2ulcs/illumos-687-webrev/
>>
>> This version makes cstyle happy.
>>
>> On Wed, May 11, 2011 at 7:10 PM, Stepan Zastupov
>> <stepan.zastupov at gmail.com> wrote:
>>> This patch introduce option -p to prevent rpcgen from generating
>>> includes with absolute paths.
>>> A few notes:
>>> 1) I use basename(3) function, which conforms to POSIX.1-2001. Is it fine?
>>> 2) I haven't documented the option since I'm not sure is '-p' a good name.
>>>
>>> http://cr.illumos.org/view/mchniy2i/illumos-687-webrev/
>
> If we look at, for example, this generated file:
>  usr/src/uts/common/idmap/idmap_xdr.c
> The generated "top matter" contains:
>  #include "/tank/ws/nfslm1/usr/src/uts/common/rpcsvc/idmap_prot.h"
>
> If I understand correctly, with the proposed -p option, this would be:
>  #include "idmap_prot.h"
>
> What I'd prefer to have in cases like this is:
>  #include "rpcsvc/idmap_prot.h"
>
> So, how about a more flexible option?  Perhaps:
>
> -X include_prefix
>    Set the prefix path used for include statements
>    in generated code.  For example:
>        rpcgen -X "rpcsvc/" ... idmap_prot.x
>    would generate includes like this:
>        #include "rpcsvc/idmap_prot.h"
>    and to get no prefix at all, use:
>        rpcgen -X "" ... idmap_prot.x
>    which would generate includes like this:
>        #include "idmap_prot.h"
>   The default (with no -X option) would be
>   the current behavior, which is to use the
>   full path as specified to find the *.x file.
>
Sounds good.

> Then, re. implementation - Why add the new args to
> all those funtions instead of adding a global option?
> That would avoid a bunch of changes.
You're right. rpcgen has too much global state already so one more can't hurt :)

>
> Factoring out the add_include function looks good.
>
> Thanks,
> Gordon
>