I think the changes file is limited by a different constraint internal to the image.
On 9/22/07, Matthew Fulmer tapplek@gmail.com wrote:
On Sat, Sep 22, 2007 at 04:08:39PM -0700, Andreas Raab wrote:
What "2GB file-addressability" limit are you referring to? There is no such thing. Files have been 64 bit for a couple of years now so unless you run some really, really old VMs that limit doesn't exists.
So, does that mean that the .changes file collapse in 3.9 was unnecessary? I thought this limit was the cause of the compresssed sources in 3.9.
-- Matthew Fulmer -- http://mtfulmer.wordpress.com/ Help improve Squeak Documentation: http://wiki.squeak.org/squeak/808
Hehe, didn't see you opened a new thread. So to maximize confusion:
Congrats to the new Magma version, great to see this living! I wonder what became of my Magma storage backend for SqueakSource - it was only a hack but *much* more scalable than any of the other backends.
- Bert -
On Sep 23, 2007, at 1:19 , Chris Muller wrote:
I think the changes file is limited by a different constraint internal to the image.
On 9/22/07, Matthew Fulmer tapplek@gmail.com wrote:
On Sat, Sep 22, 2007 at 04:08:39PM -0700, Andreas Raab wrote:
What "2GB file-addressability" limit are you referring to? There is no such thing. Files have been 64 bit for a couple of years now so unless you run some really, really old VMs that limit doesn't exists.
So, does that mean that the .changes file collapse in 3.9 was unnecessary? I thought this limit was the cause of the compresssed sources in 3.9.
-- Matthew Fulmer -- http://mtfulmer.wordpress.com/ Help improve Squeak Documentation: http://wiki.squeak.org/squeak/808
On 22-Sep-07, at 4:19 PM, Chris Muller wrote:
I think the changes file is limited by a different constraint internal to the image.
Yup; the file offset is encrypted in the tail bytes of the compiledmethod. I have made efforts to produce a clean CM design (see assorted mail on this list many times over) but I was defeated in my last attempt by a combination of the lack of time and traits not being properly supported in the image. Oh and about a hundred badly written and totally twisted (ab)uses of compiled methods.
It *could* be cleaned up. Of course it could. But it's more work than I have time for in any easily foreseeable scenario and to get maximum benefit we would have to re-do 3.9 (and 3.10) to skip the compression of the sources that would no longer be needed in order to regain the method history.
tim -- tim Rowledge; tim@rowledge.org; http://www.rowledge.org/tim Decafalon (n.): The grueling event of getting through the day consuming only things that are good for you.
On Sep 23, 2007, at 07:18 , tim Rowledge wrote:
On 22-Sep-07, at 4:19 PM, Chris Muller wrote:
I think the changes file is limited by a different constraint internal to the image.
Yup; the file offset is encrypted in the tail bytes of the compiledmethod. I have made efforts to produce a clean CM design (see assorted mail on this list many times over) but I was defeated in my last attempt by a combination of the lack of time and traits not being properly supported in the image.
Not sure what you are referring to concerning Traits? We once discussed about shared compiled methods which, compilcated your refactorings. I then modified the implementation and sent you a changeset (this was October 13, 2006).
Adrian
Oh and about a hundred badly written and totally twisted (ab)uses of compiled methods.
It *could* be cleaned up. Of course it could. But it's more work than I have time for in any easily foreseeable scenario and to get maximum benefit we would have to re-do 3.9 (and 3.10) to skip the compression of the sources that would no longer be needed in order to regain the method history.
tim
tim Rowledge; tim@rowledge.org; http://www.rowledge.org/tim Decafalon (n.): The grueling event of getting through the day consuming only things that are good for you.
On 23-Sep-07, at 3:26 AM, Adrian Lienhard wrote:
combination of the lack of time and traits not being properly supported in the image.
Not sure what you are referring to concerning Traits? We once discussed about shared compiled methods which, compilcated your refactorings. I then modified the implementation and sent you a changeset
Indeed but that doesn't make Traits fully supported by the browser (s). And as I said, that was one of several serious problems I had when trying to fix CMs.
tim -- tim Rowledge; tim@rowledge.org; http://www.rowledge.org/tim Oxymorons: Taped live
squeak-dev@lists.squeakfoundation.org