https://svn.lrde.epita.fr/svn/xrm/trunk Index: ChangeLog from SIGOURE Benoit &lt;sigoure.benoit(a)lrde.epita.fr&gt; Fix boxing of meta-if at Expression level. The new test fails because of an ambiguity. This ambiguity is a bit tricky and (IMHO) must be resolved at the grammar level. Basically we have: test1 ? test1-true : test2 ? test2-true : test2-false The ambiguity is: (test1 ? test1-true : test2) ? test2-true : test2-false -or- test1 ? test1-true : (test2 ? test2-true : test2-false) The second choice seems more natural and that's how it works in C. Obviously we can't keep this ambiguity in the grammar. * tests/xrm/amb-if-exp.xpm: New. * src/lib/xrm/pp/xrm-meta-if.str: Fix boxing of meta-if. src/lib/xrm/pp/xrm-meta-if.str | 29 +++++++++++++++++++++++++++++ tests/xrm/amb-if-exp.xpm | 14 ++++++++++++++ 2 files changed, 43 insertions(+) Index: tests/xrm/amb-if-exp.xpm --- tests/xrm/amb-if-exp.xpm (revision 0) +++ tests/xrm/amb-if-exp.xpm (revision 0) @@ -0,0 +1,14 @@ +for y from 0 to height - 1 do + module sensor + [] true -> (s'=test1 ? test1-true : test2 ? test2-true : test2-false); + endmodule +end + +// amb + +//bad: +// (test1 ? test1-true : test2) ? test2-true : test2-false + +//good: +// test1 ? test1-true : (test2 ? test2-true : test2-false) +//at least that's how it works in C and it feels more natural. Index: src/lib/xrm/pp/xrm-meta-if.str --- src/lib/xrm/pp/xrm-meta-if.str (revision 56) +++ src/lib/xrm/pp/xrm-meta-if.str (working copy) @@ -2,11 +2,16 @@ rules + /* Boxing for if statements at top level or inside modules. + * Their then-part and else-part are lists. + */ + prism-to-box: MetaIf(condition, then-part) -> V[ V is=2 [ H hs=1 [ KW["if"] ~condition KW["then"] ] ~*then-part ] KW["end"]] + where <is-list> then-part prism-to-box: MetaIf(condition, then-part, else-part) @@ -14,3 +19,27 @@ ~*then-part ] V is=2 [ KW["else"] ~*else-part ] KW["end"]] + where <is-list> then-part + ; <is-list> else-part + + /* Boxing for if statements at the Expression level. + * Their then-part and else-part are not list but instead simple elements. + * This is due to the fact that the language as no way to express + * statements nor expression sequences at Expression level. + */ + + prism-to-box: + MetaIf(condition, then-part) + -> V[ V is=2 [ H hs=1 [ KW["if"] ~condition KW["then"] ] + ~then-part ] + KW["end"]] + where <not(is-list)> then-part + + prism-to-box: + MetaIf(condition, then-part, else-part) + -> V[ V is=2 [ H hs=1 [ KW["if"] ~condition KW["then"] ] + ~then-part ] + V is=2 [ KW["else"] ~else-part ] + KW["end"]] + where <not(is-list)> then-part + ; <not(is-list)> else-part