Update HSL_MA97 headers for new version

[jfowkes]

Hi @svigerske

We're in the process of updating HSL_MA97 to v2.8.0 and in particular we have found an error in the reporting of the maximum front size statistic. This is being corrected to report the maximum number of rows in a supernode and the previously reported statistic is being reported as the maximum supernode size.

This has necessitated adding a new maxsupernode member to the ma97_info structs, using up one of the reserved integer spaces in the struct. As the info struct size remains the same I believe this change should be full backward compatible (earlier versions of HSL_MA97 will simply not have this entry set).

However, if you do think that this will be a breaking change, we are happy to drop it from the C interface.

[CLAassistant]

All committers have signed the CLA.

[svigerske]

Doesn't sound like a breaking change. It also looks like is wouldn't make any change to Ipopt, since that int that gets renamed wasn't used and will probably not be used for now.

But it would be nice to not rename the struct. The Ipopt/MA97 interface assumes that it is always called ma97_info, whether building single or double precision.

[jfowkes]

Thank you for the input, actually the struct name does not change as it gets re-defined at the start of the header (Line 47) as has always been the case:

#define ma97_info ma97_info_d

along with all the other structs and functions (e.g. ma97_control_d). Naming the struct ma97_info despite the #define was fixed quite a while ago in v2.7.1 of HSL_MA97.

[svigerske]

Ah yes. Thanks.