astmatcher-demo exposes major bugs in bclasp 0.3-test-7-8-g42f5e53
I decided to test astmatcher-demo[1] (known to be working on f7948ae i.e. tip of master) on bclasp 0.3-test-7-8-g42f5e53 (tip of newbuild at the time). First, there is a missing external symbol NEW-DEDUPLICATING-REFACTORING-TOOL (called from src/lisp/clang-tool.lsp). I assumed that NEW-REFACTORING-TOOL would be a suitable replacement, evaluated the calling function definition (C-M-x in SLIME) and moved on. Then, a series of undefined symbols (see Table 1) followed, each one "fixed" by C-M-x. After this, astmatcher-demo's GET-CLASSES fails with "Bad keyword argument" followed by a large number (for example, 69993087673325; a decimal pointer address?) which was different between bclasp instances. On further investigation, it is interesting to note that LOADing clang-tool.lsp rather than REQUIREing 'clang-tool results in more undefined symbol errors (including the external symbol REPLACEMENTS-AS-LIST). I believe this is indicative of a significant difference between evaluation and compilation in bclasp. Table 1: clang-tool.lsp symbols misreported as undefined +----------------+--------------------------------------+ | DEFUN|IDENTIFY-HINT-ENVIRONMENT | |DEFINE-CONDITION|NODE-MATCHER-AMBIGUOUS-ERROR | | DEFUN|ERROR-UNLESS-VALID-PREDICATE | | DEFCONSTANT|+NODE-MATCHERS+ | | DEFUN|IDENTIFY-NODE-MATCHER-ENVIRONMENT | | DEFCONSTANT|+TRAVERSAL-MATCHERS+ | | DEFUN|IDENTIFY-TRAVERSAL-MATCHER-ENVIRONMENT| | DEFUN|APPLICABLE-MATCHER-P | | DEFUN|MATCH-RUN | +----------------+--------------------------------------+ References [1] https://github.com/drmeister/astmatcher-demo
Most of these issues have been resolved as of 0.3-test-8-4-g653ee86, however the "Bad keyword argument" error is still extant. Regards, Georgiy On 09 Jul, Georgiy Tugai wrote:
I decided to test astmatcher-demo[1] (known to be working on f7948ae i.e. tip of master) on bclasp 0.3-test-7-8-g42f5e53 (tip of newbuild at the time).
First, there is a missing external symbol NEW-DEDUPLICATING-REFACTORING-TOOL (called from src/lisp/clang-tool.lsp). I assumed that NEW-REFACTORING-TOOL would be a suitable replacement, evaluated the calling function definition (C-M-x in SLIME) and moved on.
Then, a series of undefined symbols (see Table 1) followed, each one "fixed" by C-M-x.
After this, astmatcher-demo's GET-CLASSES fails with "Bad keyword argument" followed by a large number (for example, 69993087673325; a decimal pointer address?) which was different between bclasp instances.
On further investigation, it is interesting to note that LOADing clang-tool.lsp rather than REQUIREing 'clang-tool results in more undefined symbol errors (including the external symbol REPLACEMENTS-AS-LIST). I believe this is indicative of a significant difference between evaluation and compilation in bclasp.
Table 1: clang-tool.lsp symbols misreported as undefined +----------------+--------------------------------------+ | DEFUN|IDENTIFY-HINT-ENVIRONMENT | |DEFINE-CONDITION|NODE-MATCHER-AMBIGUOUS-ERROR | | DEFUN|ERROR-UNLESS-VALID-PREDICATE | | DEFCONSTANT|+NODE-MATCHERS+ | | DEFUN|IDENTIFY-NODE-MATCHER-ENVIRONMENT | | DEFCONSTANT|+TRAVERSAL-MATCHERS+ | | DEFUN|IDENTIFY-TRAVERSAL-MATCHER-ENVIRONMENT| | DEFUN|APPLICABLE-MATCHER-P | | DEFUN|MATCH-RUN | +----------------+--------------------------------------+
References [1] https://github.com/drmeister/astmatcher-demo
"Bad keyword argument" apparently means that a Lisp lambda-list (without &allow-other-keys) is receiving a keyword which is not in the set of known arguments. MATCH-RUN's argument signature has changed since astmatcher-demo was first written, thus the issue. I understand that Clasp is alpha software and thus unexpected major API changes are business as usual, however it'd be nice to have some backwards compatibility or graceful failure (whether a message to the effect of "use this instead", or more useful error messages; e.g. "bad keyword argument" reporting the symbol name rather than a pointer). In this particular case, backwards compatibility can be achieved quite easily, however if there are few current users of this API (other than astmatcher-demo), this would only create code cruft in Clasp. Regards, Georgiy On 10 Jul, Georgiy Tugai wrote:
Most of these issues have been resolved as of 0.3-test-8-4-g653ee86, however the "Bad keyword argument" error is still extant.
Regards, Georgiy
On 09 Jul, Georgiy Tugai wrote:
I decided to test astmatcher-demo[1] (known to be working on f7948ae i.e. tip of master) on bclasp 0.3-test-7-8-g42f5e53 (tip of newbuild at the time).
First, there is a missing external symbol NEW-DEDUPLICATING-REFACTORING-TOOL (called from src/lisp/clang-tool.lsp). I assumed that NEW-REFACTORING-TOOL would be a suitable replacement, evaluated the calling function definition (C-M-x in SLIME) and moved on.
Then, a series of undefined symbols (see Table 1) followed, each one "fixed" by C-M-x.
After this, astmatcher-demo's GET-CLASSES fails with "Bad keyword argument" followed by a large number (for example, 69993087673325; a decimal pointer address?) which was different between bclasp instances.
On further investigation, it is interesting to note that LOADing clang-tool.lsp rather than REQUIREing 'clang-tool results in more undefined symbol errors (including the external symbol REPLACEMENTS-AS-LIST). I believe this is indicative of a significant difference between evaluation and compilation in bclasp.
Table 1: clang-tool.lsp symbols misreported as undefined +----------------+--------------------------------------+ | DEFUN|IDENTIFY-HINT-ENVIRONMENT | |DEFINE-CONDITION|NODE-MATCHER-AMBIGUOUS-ERROR | | DEFUN|ERROR-UNLESS-VALID-PREDICATE | | DEFCONSTANT|+NODE-MATCHERS+ | | DEFUN|IDENTIFY-NODE-MATCHER-ENVIRONMENT | | DEFCONSTANT|+TRAVERSAL-MATCHERS+ | | DEFUN|IDENTIFY-TRAVERSAL-MATCHER-ENVIRONMENT| | DEFUN|APPLICABLE-MATCHER-P | | DEFUN|MATCH-RUN | +----------------+--------------------------------------+
References [1] https://github.com/drmeister/astmatcher-demo
participants (1)
-
Georgiy Tugai