Handling unrecoverable ABRT signals in HDF5 Library #126
Conversation
… checking" This reverts commit fc62957.
|
@anthonypapetti Thanks for contributing this. The tests pass (except for the reporter, which I expect to fail since this is a PR; need to fix that to not run for PRs). Going to merge. |
|
@selimnairb Found afterwards that this also fixes #81 , will update if I find any more that is fixed by it. |
|
@selimnairb Found that it fixes #103 #79 #93 |
| @@ -859,6 +859,7 @@ bool decodeDataQualityInfo( | |||
| else if (schemaVersion == 2) | |||
| { | |||
| //gmd:DQ_DataQuality/gmd:scope/gmd:DQ_Scope/gmd:level | |||
There was a problem hiding this comment.
Would it be better here to change BagDataQuality::scope to be a std::string instead of a char*? It would mean we wouldn't have to use delete[] manually everywhere, which might also help with bugs
| @@ -1055,6 +1056,11 @@ std::tuple<double, double> Dataset::gridToGeo( | |||
| return {x, y}; | |||
| } | |||
|
|
|||
| void handleAbrt(int signum) { | |||
There was a problem hiding this comment.
I think this runs into some problems that @selimnairb and I talked about during the last meeting, I'm working on some code now to handle it "safely" within the specification. The worry is that using the std::cerr and streams allocates, which isn't well defined behavior within an abort handler. I wonder if there's a way that we can tell aborts apart from segfaults?
Fixes #118 #117 #84 #80
This commit adds error handling using std::signal to catch SIGABRT signals from the HDF5 Library in bag_dataset.cpp and exit the program gracefully.