On Mon, Jun 25, 2012 at 6:05 PM, Sean P. DeNigris <sean@clipperadams.com> wrote:

Mariano Martinez Peck wrote

>
> Sean asked me if the package contained non-asci characters..and yes, it
> has
> widestrings directly in the literal of a method (this is on purpose).
> The method is #testWideString
>

testWideString contains wide characters. But the write is not failing inside
testWideString because zip writing is buffered in 4096 byte chunks, so it
fails at the beginning of that buffer write (in testTimestamp).
If I remove the wide char literals from TestWideString, and resave the
package, I can unzip it and file in source.st

Wow. I love this community. Pavel founds the problem. Sean debugs it and founds the real cause.
Thank you guys, you cover our back when I don't have much time :)
So...as a workaround, since the literals of a CM are visited as any other object, it is not really needed to have a test with WideStrings in the source code.
So I have just have:

testWideString

�� self assertSerializationEqualityOf: 'aString' asWideString.
�� self assertSerializationEqualityOf: (WideString
�� streamContents: [ :stream |
�� 2000 timesRepeat: [
�� stream nextPut: (256 to: 1000) atRandom asCharacter ] ] ).

and that should work :)

�

The write logic is in ZipArchiveMember>>copyDataWithCRCTo:
� � � � � � � �...
� � � � � � � �data := self readRawChunk: (4096 min: readDataRemaining).
� � � � � � � �aStream nextPutAll: data.
� � � � � � � �...

When #readRawChunk: returns a WideString, all hell breaks loose. I don't
know what the solution is, but at least we found the problem...

We should try to write a unit test for this. I will at least open an issue ;)

�

--
View this message in context: http://forum.world.st/Problem-with-Monticelo-package-and-corrupted-sourcecode-zip-tp4636458p4636561.html
Sent from the Pharo Smalltalk mailing list archive at Nabble.com.