[squeak-dev] VM Crashes when launched from home directory of user with Polish characters

> The following bugs reported and fixed in Pharo might be the same or at
> least related:
> http://code.google.com/p/pharo/issues/detail?id=513
> http://code.google.com/p/pharo/issues/detail?id=370
>
> There are more debug logs attached which might help fix it.
>
> Cheers,
> Bernhard
>
> Am 10.07.2009 um 21:22 schrieb Marcin Raczkowski:
>> When running Squeak VM from home directory of user that conatains
>> Polish characters It crashes and refuses to launch.
>> When moved to root directory of the drive, it starts but reports
>> critical error "Invalid utf8 input detected."
>>
>> Backtrace suggests its SecurityManager trying to read users home
>> directory. The same bug prevents from saving image.
>>
>> Everything works fine when launched from account which name does not
>> containing polish characters.
>>
>> Someone suggested problem might be solved in E-Toys.
>>
>> Additional Information:
>>
>> OS: Vista (Polish version)
>> Fiesystem: NTFS
>> Squek downloded from
>> http://ftp.squeak.org/3.10/win/Squeak3.10.2-7179-win32.zip
>>
>> Ticket url: http://bugs.squeak.org/view.php?id=7365
>> Wasn't sure under what category fill it, since it seems to affect both
>> VM and system image
>>
>
>
> ------------------------------------------------------------------------
>
>

> 2009/7/11 Andreas Raab <[hidden email]>:
>> At Qwaq, we've changed our utf8ToSqueak version to assume that the input is
>> in Latin-1 when "invalid input" is encountered. That's because we've never
>> found a situation where one wouldn't be better off to silently accept the
>> input (regardless of whether it's actually UTF-8, Latin-1 or whatever else)
>> instead of blowing up. I wonder what people think about that?
>
> It is just a way of masking bugs. At that point the only thing you
> know is that the input is not utf-8. Coming up with some random
> encoding is likely to produce corrupted data. For Seaside this would
> be horrible.

> Philippe Marschall wrote:
>>
>> 2009/7/11 Andreas Raab <[hidden email]>:
>>>
>>> At Qwaq, we've changed our utf8ToSqueak version to assume that the input
>>> is
>>> in Latin-1 when "invalid input" is encountered. That's because we've
>>> never
>>> found a situation where one wouldn't be better off to silently accept the
>>> input (regardless of whether it's actually UTF-8, Latin-1 or whatever
>>> else)
>>> instead of blowing up. I wonder what people think about that?
>>
>> It is just a way of masking bugs. At that point the only thing you
>> know is that the input is not utf-8. Coming up with some random
>> encoding is likely to produce corrupted data. For Seaside this would
>> be horrible.
>
> Fair enough, that's why I'm asking ;-) But I'm curious, have you *ever*
> actually encountered the situation where the error about "invalid utf8
> input" has helped you?