Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
4 check-ins that include changes to files matching 'modules/pt/tests/data/gr/fail-sym-sequence-container-res/*'
|
2014-10-20
| ||
| 23:45 | Merged trunk into the long-dormant branch for updating the C pieces to critcl 3. check-in: 3a72abc855 user: andreask tags: update-to-critcl3 | |
|
2014-07-21
| ||
| 18:57 | Merged work on the parsetools into trunk. Very extended testsuite, and lots of bugfixes, especially in the area of error handling. Also fixes in the C runtime and generated code where long/void* size mismatches caused crashing. check-in: 598ddec2b2 user: andreask tags: trunk | |
|
2014-06-10
| ||
| 20:15 | Disabled the i_error_nonterminal instruction in all implementations of the RDE runtime (Tcl snit|oo, Critcl). This means that at the moment the error state of a pt-made parser will only contain messages with references to expected terminal symbols. Non-terminal symbols will not occur. This may change in the future, when it is worked out how to handle them. Updated the testsuite, as less as there is, to match this. check-in: 1afbe0b080 user: andreask tags: pt-work | |
| 07:28 | Extended coverage, simple sequence accessed through a symbol. This already shows an issue with error handling when using a plain sequence. An error at the beginning is shown as symbol error (n N). A later error in the sequence is shown as terminal error (t c). The issue is in the "i_error_nonterminal" and how it merges. The conflicting goals are "report error at farthest position" vs "report a highlevel error (symbol)". Currently resolved in favor of "farthest", test case uses "symbol" instead, to demonstrate the conflict. If we wish to have both farthest and nice we have to redo the data structures to accomodate storage of errors at different locations, which are related. This requires more thoughts. check-in: 7c2bac860c user: aku tags: pt-work | |