scda: Write and read a file header

[tim-griesbach]

scda: Write and read a file header

This PR adds the functionality to read and write a scda file header, i.e. implementing in particular the functions fopen_write and fopen_read according to the scda API introduced in #132 and #156.

A list of some of the functionalities that are implemented in this PR is

• fopen_read and fopen_write,
• helper functions that pad data according to the scda convention defined in the scda preprint,
• helper functions that check if data is correctly padded and that extract the actual data,
• complete error handling for fopen_read and fopen_write; including MPI count error management,
• ferror_string to convert error codes to error strings,
• fuzzy error checking to increase the code coverage of the error management,
• a basic test program that currently aborts on errors and offers fuzzy error testing by command line parameters and
• the non-static function sc_scda_is_success to check if an error code encodes success.

[cburstedde]

When merging the latest develop branch, will the CI go through?

[tim-griesbach]

When merging the latest develop branch, will the CI go through?

Yes, I think so. I merged the latest develop branch into this PR and the CI looks good to me.

[cburstedde]

Thanks! I'll be merging this one now.

For future development: Would it be more intuitive to rename the function to sc_scda_ferror_success (or something else -- open for suggestions) and to return MPI_SUCCESS on success and some MPI error otherwise?

[tim-griesbach]

I think that a function sc_scda_ferror_success that returns MPI error codes may be unexpected to the user and we also have the sc_scda_ret_t value SC_SCDA_FERR_SUCCESS that may be named too similar.
Do you think it is sensible to name the function sc_scda_ferror_is_success to be consistent to the other scda function names but still use the semantic of mapping an error code to a Boolean?

[cburstedde]

+/** Check if a scda_errorcode_t encodes success. + * + * \param [in] errorcode An errcode that is output by a sc_scda function. + * \return True if \b errcode encodes success and + * false otherwise. + */ +int sc_scda_is_success (sc_scda_ferror_t errorcode); I think that a function `sc_scda_ferror_success` that returns MPI error codes may be unexpected to the user and we also have the `sc_scda_ret_t` value `SC_SCDA_FERR_SUCCESS` that may be named too similar.

Good point.

Do you think it is sensible to name the function `sc_scda_ferror_is_success` to be consistent to the other scda function names but still use the semantic of mapping an error code to a Boolean?

I like this proposal!