[PATCH] BBHandler: Ensure parser state engine is correctly reset for new parsing

[PATCH] BBHandler: Ensure parser state engine is correctly reset for new parsing

[Richard Purdie]

The classes variable in particular could be in an odd state if parsing of a
previous recipe had abruptly ended.

Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
---
diff --git a/bitbake/lib/bb/parse/parse_py/BBHandler.py b/bitbake/lib/bb/parse/parse_py/BBHandler.py
index 2e0647b..c4e579d 100644
--- a/bitbake/lib/bb/parse/parse_py/BBHandler.py
+++ b/bitbake/lib/bb/parse/parse_py/BBHandler.py
@@ -107,14 +107,15 @@ def get_statements(filename, absolute_filename, base_name):
         return statements
 
 def handle(fn, d, include):
-    global __func_start_regexp__, __inherit_regexp__, __export_func_regexp__, __addtask_regexp__, __addhandler_regexp__, __infunc__, __body__, __residue__
+    global __func_start_regexp__, __inherit_regexp__, __export_func_regexp__, __addtask_regexp__, __addhandler_regexp__, __infunc__, __body__, __residue__, __inpython__, classes
     __body__ = []
     __infunc__ = ""
     __classname__ = ""
     __residue__ = []
-
+    __inpython__ = False
 
     if include == 0:
+        classes = [ None, ]
         logger.debug(2, "BB %s: handle(data)", fn)
     else:
         logger.debug(2, "BB %s: handle(data, include)", fn)

Re: [PATCH] BBHandler: Ensure parser state engine is correctly reset for new parsing

[Richard Purdie]

On Fri, 2012-12-07 at 13:10 +0000, Richard Purdie wrote:
> The classes variable in particular could be in an odd state if parsing of a
> previous recipe had abruptly ended.
> 
> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
> ---
> diff --git a/bitbake/lib/bb/parse/parse_py/BBHandler.py b/bitbake/lib/bb/parse/parse_py/BBHandler.py
> index 2e0647b..c4e579d 100644
> --- a/bitbake/lib/bb/parse/parse_py/BBHandler.py
> +++ b/bitbake/lib/bb/parse/parse_py/BBHandler.py
> @@ -107,14 +107,15 @@ def get_statements(filename, absolute_filename, base_name):
>          return statements
>  
>  def handle(fn, d, include):
> -    global __func_start_regexp__, __inherit_regexp__, __export_func_regexp__, __addtask_regexp__, __addhandler_regexp__, __infunc__, __body__, __residue__
> +    global __func_start_regexp__, __inherit_regexp__, __export_func_regexp__, __addtask_regexp__, __addhandler_regexp__, __infunc__, __body__, __residue__, __inpython__, classes
>      __body__ = []
>      __infunc__ = ""
>      __classname__ = ""
>      __residue__ = []
> -
> +    __inpython__ = False
>  
>      if include == 0:
> +        classes = [ None, ]
>          logger.debug(2, "BB %s: handle(data)", fn)
>      else:
>          logger.debug(2, "BB %s: handle(data, include)", fn)

This patch doesn't work. To cut a long story short, "classes" is getting
cached inside the ast statement cache and shouldn't be. This is likely
the source of the problems.

Trying to fix that leads to a voyage of discovery of other breakage :(

Its likely going to end up with the EXPORT_FUNCTION implementation
needing rewriting, or at least how/where its getting its class lists
from.

Cheers,

Richard