Ticket #1286 (new bug)
method overload in pir subclass of pmc pccmethod leaves object undefined
Reported by: | particle | Owned by: | |
---|---|---|---|
Priority: | normal | Milestone: | |
Component: | library | Version: | trunk |
Severity: | medium | Keywords: | |
Cc: | Language: | ||
Patch status: | Platform: |
Description
while creating a subclass of Exporter to test inter-language exports, i stumbled across a bug that i can't seem to fix.
i've created a subclass of Exporter called MyExporter. MyExporter has an overridden 'add_global' method, which adds a '&' sigil to each global specified by the user and calls the method in the superclass.
i've also created a high-level language that uses the '&' sigil on subs, and i import and aim to call the hll sub. however, the code never finishes 'add_method', as it segfaults inside the supermethod with an undefined 'exp' object. that is, from src/pmc/exporter.pmc, line 260:
Parrot_Exporter *exp = PARROT_EXPORTER(SELF);
after this line, 'exp' is NULL, but 'SELF' (or after macro expansion 'pmc') is not. this is not the case when 'add_method' has not been overridden. here's the pir code that exposes the bug:
.sub 'main' :main .local pmc exp, ns
exp = new 'MyExporter'
ns = get_root_namespace foo? exp.'source'( ns )
exp.'add_global'('bar')
say "NEVER REACHES HERE"
exp.'import'()
'&bar'() ## THIS WILL SUCCEED! 'bar'() ## THIS WILL FAIL! ## THAT'S NOT WHAT I WANT!!! .end
## create a subclass of Exporter that overrides 'add_global' ## it adds a '&' sigil to each requested global .namespace MyExporter? .sub 'onload' :load :init :anon .local pmc class class = subclass 'Exporter', 'MyExporter' .return () .end
.sub 'add_global' :method .param string glb :optional .param int has_glb :opt_flag
.local pmc args, super args = new .Hash super = new .Super, self
unless has_glb goto no_global glb = concat '&', glb .return super.'add_global'( glb ) no_global: .return super.'add_global'() .end
# create a high-level language that uses '&' sigil for sub names .HLL 'foo', .sub '&bar' say 'hello, world!' .end
note that switching from a tail method call to a regular method call has no effect.
also note that my code also exposes a problem with the current Exporter implementation: the imported sub name does not match the requested name. i have yet to address this as the current bug is blocking progress.
~jerry