On Sun, Jul 06, 2014 at 12:47:25PM -0400, David T. Lewis wrote:
On Sun, Jul 06, 2014 at 12:23:11PM -0400, David T. Lewis wrote:
On Sun, Jul 06, 2014 at 10:34:15AM +0200, Nicolas Cellier wrote:
2014-07-04 0:47 GMT+02:00 Nicolas Cellier < nicolas.cellier.aka.nice@gmail.com>:
2014-07-04 0:09 GMT+02:00 Eliot Miranda eliot.miranda@gmail.com:
On Thu, Jul 3, 2014 at 12:33 PM, Nicolas Cellier < nicolas.cellier.aka.nice@gmail.com> wrote:
The bug is repeatable, i simply have to execute this snippet with my test file:
(FileStream fileNamed: 'snapshot.bin') binary contentsOfEntireFile asString zipped.
The file is too big for an attachment here - 7.3 Mbytes - or 1.7 Mbytes gzipped by external tool. If someone can suggest an upload site, or want it by mail, just ask.
You're welcome to put it on ftp.mirandanamda.org, cogftpuser, pw cogging with 0's & 1's.
done, pollution of your site engaged... Thanks Eliot!
I inquired a bit more about this bug. An important clue is that it does not happen in Pharo3.0!
But Pharo3.0 did not fundamentally change compression (except some FileSystem related changes, separation of CRC stuff in another package, some other cosmetic changes like replacing some self assert: ... by [...] assert and a potential bug in DeflateStream>>nextPutAll:startingAt: introduced by CamilleTeruel). Squeak behavior is presumably not due to a Compression change.
Neither is it a VM problem, the bugs still shows up when running the Squeak image with Pharo VM.
Yes, it is definite a problem in the Squeak trunk image.
So the difference lies somewhere else: in our beloved Stream. I remembered a recent change of Eliot related to handling of Stream created with on:from:to: (Collections-eem.567). Reverting those changes just make the snippet pass!
Ah ah! At the time, i found the change of Eliot quite minimal and nice. But I wish I had time to analyze this small innocent change deeper. Indeed, I remembered I previously broke my teeth on this one 2 years or so ago, and preferred to adopt another strategy: avoid using on:from:to: and ReadWriteStream as much as possible. Why? because when analyzing the usage of inst.var. in Stream hierarchy, it gave me headaches, some subclass are just ignoring superclass variables, or worse are bending the semantics of superclass variables. I came to the conclusion that I could not decently master a change of on:from:to:
I can not confirm this. I have an image in which I can reliably reproduce the failure in writing the MCZ to a file. I tried reverting the methods that were introduced in Collections-eem.567 and I am still getting the same failure.
Dave
Oops, I spoke too soon. Indeed the problem appears as of Collections-eem.567, and does not appear in Collections-nice.566.
So I am confirming Nicolas' observations ... now to find the bug :)
Dave
This seems to be a rather tricky bug, and I don't think any of us knows the cause right now. In case anyone else wants to have a look at it, here is a recipe for reproducing the bug in a clean 4.5 image:
- Start with a clean Squeak 4.5 image in and empty working directory. ftp://ftp.squeak.org/4.5/Squeak4.5-13680.zip
- Open the image, do not update it.
- Open an MC browser and add the http://source.squeak.org/trunk repository for the Help-Squeak-Project package and Collections package.
- Load Help-Squeak-Project-kfr.18 from trunk.
- Load Collections-eem.567 from trunk.
- Open a browser on class SqueakToolsDebuggerHelp, and delete the class side method #showDebuggerMenuForm.
- Remove the http://source.squeak.org/trunk repository form the Help-Squeak-Project package in the MC browser.
- Highlight your package-cache repository and try to save Help-Squeak-project to the package-cache repository.
- Enter author initials 'dtl'.
- For the package comment, enter 'Remove unreferenced method'.
- Package save will fail part way through saving the MCZ.
- To reproduce the failure in this image, do the following in a workspace:
mcv := MCVersion allInstances detect: [:e | e name = 'a MCVersion(Help-Squeak-Project-dtl.19)']. [f := FileStream fileNamed: 'junk.mcz'. mcv fileOutOn: f] ensure: [f close].
Dave