Hi Guys,
source.squeak.org/trunk's Morphic-eem.740(cmm.739).mcd is empty. How do we fix it?
On a related note is there any way to get the updater to download full packages, avoiding the diffy versions? e.g a preference?
On 07.08.2014, at 19:52, Eliot Miranda eliot.miranda@gmail.com wrote:
Hi Guys,
source.squeak.org/trunk's Morphic-eem.740(cmm.739).mcd is empty. How do we fix it?
Not empty, but broken:
bert$ unzip -t Morphic-eem.740(cmm.739).mcd Archive: Morphic-eem.740(cmm.739).mcd testing: package OK testing: version OK testing: base OK testing: old/source.st error: invalid compressed data to inflate testing: new/source.st error: invalid compressed data to inflate testing: patch.bin OK
Since the meta data and patch.bin are okay I'd expect this to load anyway ...
I remember seeing messages about broken ZIP files. Was there a fix? Did someone fix the squeaksource image, too? The MCDs are generated on the server.
On a related note is there any way to get the updater to download full packages, avoiding the diffy versions? e.g a preference?
No pref, but you can comment out the diffy part in MCConfiguration>>versionNamed:for:from:.
- Bert -
On Mon, Aug 11, 2014 at 8:49 AM, Bert Freudenberg bert@freudenbergs.de wrote:
On 07.08.2014, at 19:52, Eliot Miranda eliot.miranda@gmail.com wrote:
Hi Guys,
source.squeak.org/trunk's Morphic-eem.740(cmm.739).mcd is empty.How do we fix it?
Not empty, but broken:
bert$ unzip -t Morphic-eem.740(cmm.739).mcd Archive: Morphic-eem.740(cmm.739).mcd testing: package OK testing: version OK testing: base OK testing: old/source.st error: invalid compressed data to inflate testing: new/source.st error: invalid compressed data to inflate testing: patch.bin OK
Since the meta data and patch.bin are okay I'd expect this to load anyway ...
It does now. I don't understand this stuff well enough to know quite what has happened :-(.
I remember seeing messages about broken ZIP files. Was there a fix? Did someone fix the squeaksource image, too? The MCDs are generated on the server.
There was an issue with the plugin that meant that adding inst vars to WriteStream and/or ReadStream would cause the plugin primitives to fail, falling back on the image code. There is an as-yet-undiagnosed bug that we suspect is to do with wide characters in the image code that causes it to fail to compress some sources (I suspect long sources that contain wide characters). Hence when I fixed (aStream on: coll from: nGreaterThanOne to: limit) contents by adding inst vars to WriteStream and ReadStream the image-level bug was provoked.
I've fixed the plugin in recent Cog VMs and IIRC David has folded the fix into recent interpreter VMs. But that still leaves the image code to debug.
On a related note is there any way to get the updater to download full packages, avoiding the diffy versions? e.g a preference?
No pref, but you can comment out the diffy part in MCConfiguration>>versionNamed:for:from:.
Ah, OK. Mahalo! Any objections if I add a preference?
On 11.08.2014, at 20:06, Eliot Miranda eliot.miranda@gmail.com wrote:
On Mon, Aug 11, 2014 at 8:49 AM, Bert Freudenberg bert@freudenbergs.de wrote:
On a related note is there any way to get the updater to download full packages, avoiding the diffy versions? e.g a preference?
No pref, but you can comment out the diffy part in MCConfiguration>>versionNamed:for:from:.
Ah, OK. Mahalo! Any objections if I add a preference?
Well, I generally dislike proliferating preferences. The code is supposed to automatically fall back to full versions if the more efficient diffy version doesn't load. If that fallback doesn't work I'd rather fix that problem rather than avoiding it with a preference. The diffs are much smaller than full MCDs, so for server performance and conserving bandwidth they are preferable.
- Bert -
On Mon, Aug 11, 2014 at 11:41 AM, Bert Freudenberg bert@freudenbergs.de wrote:
On 11.08.2014, at 20:06, Eliot Miranda eliot.miranda@gmail.com wrote:
On Mon, Aug 11, 2014 at 8:49 AM, Bert Freudenberg bert@freudenbergs.de
wrote:
On a related note is there any way to get the updater to download
full packages, avoiding the diffy versions? e.g a preference?
No pref, but you can comment out the diffy part in
MCConfiguration>>versionNamed:for:from:.
Ah, OK. Mahalo! Any objections if I add a preference?
Well, I generally dislike proliferating preferences. The code is supposed to automatically fall back to full versions if the more efficient diffy version doesn't load. If that fallback doesn't work I'd rather fix that problem rather than avoiding it with a preference. The diffs are much smaller than full MCDs, so for server performance and conserving bandwidth they are preferable.
OK, makes sense.
On a related note is there any way to get the updater to download full packages, avoiding the diffy versions? e.g a preference?
No pref, but you can comment out the diffy part in MCConfiguration>>versionNamed:for:from:.
Ah, OK. Mahalo! Any objections if I add a preference?
Wouldn't it be better for mcd's to just work than add a preference just to get around a bug?
On Mon, Aug 11, 2014 at 12:36 PM, Chris Muller ma.chris.m@gmail.com wrote:
On a related note is there any way to get the updater to download full packages, avoiding the diffy versions? e.g a preference?
No pref, but you can comment out the diffy part in MCConfiguration>>versionNamed:for:from:.
Ah, OK. Mahalo! Any objections if I add a preference?
Wouldn't it be better for mcd's to just work than add a preference just to get around a bug?
Sure, but when one is struggling to get a bootstrap working fixing bugs from elsewhere that just get in the way is not a viable option. I just wanted to rescue the Spur bootstrap. Anyway, looks like nothing needs to be done. It appears to be working now, and I'm happy to leave that a mystery for now.
I wasn't expecting you to fix the bug, just simply thought Bert's solution (code tweak) would be sufficient for you until we can track it down. I assume you already have several code tweaks you make to your images to customize them for your own purpose (doesn't everyone?). so just one more temporarily.. I, for one, hope it does not remain a mystery much longer.
On Mon, Aug 11, 2014 at 4:08 PM, Eliot Miranda eliot.miranda@gmail.com wrote:
On Mon, Aug 11, 2014 at 12:36 PM, Chris Muller ma.chris.m@gmail.com wrote:
On a related note is there any way to get the updater to download
full
packages, avoiding the diffy versions? e.g a preference?
No pref, but you can comment out the diffy part in MCConfiguration>>versionNamed:for:from:.
Ah, OK. Mahalo! Any objections if I add a preference?
Wouldn't it be better for mcd's to just work than add a preference just to get around a bug?
Sure, but when one is struggling to get a bootstrap working fixing bugs from elsewhere that just get in the way is not a viable option. I just wanted to rescue the Spur bootstrap. Anyway, looks like nothing needs to be done. It appears to be working now, and I'm happy to leave that a mystery for now. -- best, Eliot
Dear fellow Squeakers,
Here we go again. When I try to update the latest Squeak5.1-15113.image from http://ftp.squeak.org/5.1alpha using the latest CogSpur.r3684 on OS X with an empty package-cache I run into the Warning "About to serialize an empty diffy version.“ This time with SMLoader-cmm.85(ul.83).mcd. I wonder if I am the only one.
Cheers, Bernhard
Am 07.08.2014 um 19:52 schrieb Eliot Miranda eliot.miranda@gmail.com:
Hi Guys,
source.squeak.org/trunk's Morphic-eem.740(cmm.739).mcd is empty. How do we fix it?On a related note is there any way to get the updater to download full packages, avoiding the diffy versions? e.g a preference?
best, Eliot
In this case the code in ul-83 and cmm-85 is the same, so the diff is expected to be empty. I wonder why is there a warning in such cases.
Levente
On Sun, 24 Apr 2016, Bernhard Pieber wrote:
Dear fellow Squeakers,
Here we go again. When I try to update the latest Squeak5.1-15113.image from http://ftp.squeak.org/5.1alpha using the latest CogSpur.r3684 on OS X with an empty package-cache I run into the Warning "About to serialize an empty diffy version.“ This time with SMLoader-cmm.85(ul.83).mcd. I wonder if I am the only one.
Cheers, Bernhard
Am 07.08.2014 um 19:52 schrieb Eliot Miranda eliot.miranda@gmail.com:
Hi Guys,
source.squeak.org/trunk's Morphic-eem.740(cmm.739).mcd is empty. How do we fix it?On a related note is there any way to get the updater to download full packages, avoiding the diffy versions? e.g a preference?
best, Eliot
ul.83 is not a direct ancestor of cmm.85, so I would like to know what code is creating a mcd between those two and why..
The warning about serializing empty mcz is there because we never found the bug which was serializing empty mcz's. This warning is help us identify when and why it occurs, and also to prevent the bug from silently storing an empty mcz into the repository, possibly losing code and/or causing harm to trunk.
On Sun, Apr 24, 2016 at 10:12 AM, Levente Uzonyi leves@caesar.elte.hu wrote:
In this case the code in ul-83 and cmm-85 is the same, so the diff is expected to be empty. I wonder why is there a warning in such cases.
Levente
On Sun, 24 Apr 2016, Bernhard Pieber wrote:
Dear fellow Squeakers,
Here we go again. When I try to update the latest Squeak5.1-15113.image from http://ftp.squeak.org/5.1alpha using the latest CogSpur.r3684 on OS X with an empty package-cache I run into the Warning "About to serialize an empty diffy version.“ This time with SMLoader-cmm.85(ul.83).mcd. I wonder if I am the only one.
Cheers, Bernhard
Am 07.08.2014 um 19:52 schrieb Eliot Miranda eliot.miranda@gmail.com:
Hi Guys,
source.squeak.org/trunk's Morphic-eem.740(cmm.739).mcd is empty. Howdo we fix it?
On a related note is there any way to get the updater to download full packages, avoiding the diffy versions? e.g a preference? -- best, Eliot
On Sun, 24 Apr 2016, Chris Muller wrote:
ul.83 is not a direct ancestor of cmm.85, so I would like to know what code is creating a mcd between those two and why..
It's like that because you reverted the changes of dtl.84.
The warning about serializing empty mcz is there because we never found the bug which was serializing empty mcz's. This warning is help us identify when and why it occurs, and also to prevent the bug from silently storing an empty mcz into the repository, possibly losing code and/or causing harm to trunk.
But in this case the mcd has to be empty, because the code is identical, so the warning is nothing but noise and a thing to block the automatic update.
Levente
On Sun, Apr 24, 2016 at 10:12 AM, Levente Uzonyi leves@caesar.elte.hu wrote:
In this case the code in ul-83 and cmm-85 is the same, so the diff is expected to be empty. I wonder why is there a warning in such cases.
Levente
On Sun, 24 Apr 2016, Bernhard Pieber wrote:
Dear fellow Squeakers,
Here we go again. When I try to update the latest Squeak5.1-15113.image from http://ftp.squeak.org/5.1alpha using the latest CogSpur.r3684 on OS X with an empty package-cache I run into the Warning "About to serialize an empty diffy version.“ This time with SMLoader-cmm.85(ul.83).mcd. I wonder if I am the only one.
Cheers, Bernhard
Am 07.08.2014 um 19:52 schrieb Eliot Miranda eliot.miranda@gmail.com:
Hi Guys,
source.squeak.org/trunk's Morphic-eem.740(cmm.739).mcd is empty. Howdo we fix it?
On a related note is there any way to get the updater to download full packages, avoiding the diffy versions? e.g a preference? -- best, Eliot
I fixed this case by replacing update-eem.333 with update-cmm.333 which loads dtl.84 in the interim. This should resolve Bernards issue.
On Sun, Apr 24, 2016 at 1:36 PM, Levente Uzonyi leves@caesar.elte.hu wrote:
On Sun, 24 Apr 2016, Chris Muller wrote:
ul.83 is not a direct ancestor of cmm.85, so I would like to know what code is creating a mcd between those two and why..
It's like that because you reverted the changes of dtl.84.
The warning about serializing empty mcz is there because we never found the bug which was serializing empty mcz's. This warning is help us identify when and why it occurs, and also to prevent the bug from silently storing an empty mcz into the repository, possibly losing code and/or causing harm to trunk.
But in this case the mcd has to be empty, because the code is identical, so the warning is nothing but noise and a thing to block the automatic update.
Levente
On Sun, Apr 24, 2016 at 10:12 AM, Levente Uzonyi leves@caesar.elte.hu wrote:
In this case the code in ul-83 and cmm-85 is the same, so the diff is expected to be empty. I wonder why is there a warning in such cases.
Levente
On Sun, 24 Apr 2016, Bernhard Pieber wrote:
Dear fellow Squeakers,
Here we go again. When I try to update the latest Squeak5.1-15113.image from http://ftp.squeak.org/5.1alpha using the latest CogSpur.r3684 on OS X with an empty package-cache I run into the Warning "About to serialize an empty diffy version.“ This time with SMLoader-cmm.85(ul.83).mcd. I wonder if I am the only one.
Cheers, Bernhard
Am 07.08.2014 um 19:52 schrieb Eliot Miranda eliot.miranda@gmail.com:
Hi Guys,
source.squeak.org/trunk's Morphic-eem.740(cmm.739).mcd is empty.How do we fix it?
On a related note is there any way to get the updater to download full packages, avoiding the diffy versions? e.g a preference? -- best, Eliot
squeak-dev@lists.squeakfoundation.org
-
Bernhard Pieber -
Bert Freudenberg -
Chris Muller -
Chris Muller -
Eliot Miranda -
Levente Uzonyi