Hi Dana IMHO this is a minor problem when it comes to threads and when it comes to the HDF5 library robustness overall. When I allowed threads in HDF5 (now I coordinate everything outside because it was a nightmare) I had really big problems because of the logic that restores the state of the library after an error. If everything goes as it has been carefully designed, it might work. But when an error occurs, and threeads multiply the chances, the cleanup logic is just so flawed. But this is a huge problem also with the single thread, only with multiple threads it has higher chances. To see an example of what I mean and with which I am currently having trouble, try saving a file on an almost full USB stick.
m2c -- ds 2014-03-24 19:41 GMT+01:00 Dana Robinson <[email protected]>: > Hi all, > > > > This is another (very) brief RFC that outlines how we will address some > long-standing issues with thread-local storage in the Windows thread-safe > code. Although no new public functionality will be exposed, the information > may be of interest and I'd be interested in hearing from any static + > thread-safe Windows users. > > > > The new function will appear in HDF5 1.8.13, which will be released in May. > > > > If anyone has any comments or concerns, please let us know. > > > > Cheers, > > > > Dana > > > > _______________________________________________ > Hdf-forum is for HDF software users discussion. > [email protected] > > http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org > >
_______________________________________________ Hdf-forum is for HDF software users discussion. [email protected] http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
