Re: [Swig-user] Change in handling of nested structs from 1.3.36?

[William S F. <ws...@fu...>]

Solomon Gibbs wrote:
> I have a rather nasty struct in the code I'm wrapping. Though version 
> 1.3.36, swig has always consumed it without complaining. However, I 
> recently downloaded 1.3.39 and now I get a syntax error at the line 
> corresponding to the end of the "meta_size" struct (below). The 
> preprocess result from running swig with -E looks like more or less what 
> I would expect.
> 
> Does anyone have any idea what has changed, or how to figure out I am 
> doing wrong? I've tried a couple of the --debug-* options, but the parse 
> seems to stop before they can have any effect.
> 
> Any help would be appreciated.
> 
> The following struct doesn't seem to work under 1.3.38 or 1.3.39. I 
> haven't checked 1.3.37.
> 
> typedef struct H5O_info_t {
>     unsigned long     fileno;
>     haddr_t         addr;
>     H5O_type_t         type;
>     unsigned         rc;
>     time_t        atime;
>     time_t        mtime;
>     time_t        ctime;
>     time_t        btime;
>     hsize_t         num_attrs;
>     struct {
>         unsigned version;
>         unsigned nmesgs;
>         unsigned nchunks;
>         unsigned flags;
>         struct {
>             hsize_t total;
>             hsize_t meta;
>             hsize_t mesg;
>             hsize_t free;
>         } space;
>         struct {
>             uint64_t present;
>             uint64_t shared;
>         } mesg;
>     } hdr;
>     struct {
>         H5_ih_info_t obj;
>         H5_ih_info_t attr;
>     } meta_size;
> } H5O_info_t;
> 
I don't get any error using this struct and I don't see any problems 
using valgrind on SWIG for this struct. I think the best thing is for 
you to provide the exact commandline options you are passing to SWIG and 
provide a complete standalone interface file that shows the problem. 
Also which operating system, compiler and compile options you used to 
compile SWIG.

William