[Scons-dev] Bug in env.Glob() - how do I report?

[Vasily]

Hi Dirk,

Thanks for the reply!

I get this running the SConstruct I attached:

$> scons .
scons: Reading SConscript files ...
TypeError: Tried to lookup Dir 'some_entry' as a File.:
  File "C:\Vass\Work\Temp\scons_glob_bug\SConstruct", line 22:
    d2[0].srcnode()
  File "C:\Python27\scons-2.1.0\SCons\Node\FS.py", line 743:
    srcnode.must_be_same(self.__class__)
  File "C:\Python27\scons-2.1.0\SCons\Node\FS.py", line 626:
    (self.__class__.__name__, self.path, klass.__name__))

Note: I've checked it agains SCons 2.1.0 (pretty old, but that's what we
use internally right now, maybe it's already fixed?..)

As for calling disambiguate() out of place - I understand that. Actual
issue I stubmled upon was much, much more complex, this was just making it
as small as possible.
Anyway, after disambiguating d1 it gets correctly set as Dir.

I can try to re-write a reproducer the way it would be calling stuff in a
build phase, but I doubt it will change a lot - the original issue (before
the simplifying to this SConstruct) was reproduced when srcnode() was
called after all targets were built, via python's atexit mechanism.

Thanks,
Vasily

2014-12-15 19:29 GMT+03:00 Dirk Bächle <tshortik at gmx.de>:
>
> Hi Vasily,
>
> first of all "Thanks a lot" for asking first, before simply going ahead
> and submitting an issue. Asking on the User mailing list is the preferred
> method for finding out whether an odd behaviour is really a bug or not.
>
> On 15.12.2014 14:12, Vasily wrote:
>
>> Hi everybody,
>>
>> I've stubmled upon the bug in env.Glob() that loses FS.Entry actual class
>> when working with several VariantDir's coming from one source.
>>
>> Small reproducer attached.
>>
>>
> I tried to run your SConstruct on my machine, but I get an "dependency
> cycle" error. Can you provide a little bit more description about what
> output you would expect, and what you see instead? That'll help me (and
> other users) understand better what your actual problem is. I'm currently
> not able to reproduce your example on my side, so I can't debug it either.
>
> Please also note that you're using "disambiguate" on a Node, that might
> not yet exist physically. It gets created later, when the Command() builder
> is run during the actual "build phase"....
>
>  Where should I submit the issue? scons.tigris.org <
>> http://scons.tigris.org> looks a bit outdated, should I submit a bug on
>> bitbucket.org <http://bitbucket.org>?
>>
>>
> The Tigris bug tracker is still the correct place to enter issues, but
> first let's see if we can identify this as bug at all.
>
> Best regards,
>
> Dirk
>
> _______________________________________________
> Scons-dev mailing list
> Scons-dev at scons.org
> https://pairlist2.pair.net/mailman/listinfo/scons-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://pairlist2.pair.net/pipermail/scons-dev/attachments/20141215/bd03a567/attachment.html>