Hi all
ok I created a ScriptLoader package which contains scripts.... and now each time I push something in 39 I update the script and test that I can load everything from the latest ***essential image*** = that could not be created with MC. This way we can accumulate changes and from time to time publish a new image.
Script 1, 2, 3 loaded perfectly (modulo the merge making packages dirty) from 6693 + 274 MC So I will continue like that.
Stef
Am 29.09.2005 um 21:19 schrieb stéphane ducasse:
Hi all
ok I created a ScriptLoader package which contains scripts.... and now each time I push something in 39 I update the script and test that I can load everything from the latest ***essential image*** = that could not be created with MC. This way we can accumulate changes and from time to time publish a new image.
Script 1, 2, 3 loaded perfectly (modulo the merge making packages dirty) from 6693 + 274 MC So I will continue like that.
One strange thing: The image is *huge* after running the script: 3393md4 is over 20MB (MC cache deleted), 3393 was only around 13MB.
I deleted all the changesets but even then it's 18MB. What could take that amount of space?
Marcus
Monticello's caching? The magic incantation MCCacheRepository flushAllCaches (from the back of my head, correct me if I'm wrong) usually shrinks the image by quite a bit...
On 10/5/05, Marcus Denker denker@iam.unibe.ch wrote:
Am 29.09.2005 um 21:19 schrieb stéphane ducasse:
Hi all
ok I created a ScriptLoader package which contains scripts.... and now each time I push something in 39 I update the script and test that I can load everything from the latest ***essential image*** = that could not be created with MC. This way we can accumulate changes and from time to time publish a new image.
Script 1, 2, 3 loaded perfectly (modulo the merge making packages dirty) from 6693 + 274 MC So I will continue like that.
One strange thing: The image is *huge* after running the script: 3393md4 is over 20MB (MC cache deleted), 3393 was only around 13MB.
I deleted all the changesets but even then it's 18MB. What could take that amount of space?
Marcus
Am 05.10.2005 um 12:27 schrieb Marcus Denker:
Am 05.10.2005 um 11:48 schrieb Cees De Groot:
Monticello's caching? The magic incantation MCCacheRepository flushAllCaches (from the back of my head, correct me if I'm wrong) usually shrinks the image by quite a bit...
No, I've already cleaned the mc cache.
It could be that someting is wrong with the cleaning... I just did a flush in my newly build work-image and it said: "3 versions flushed, 344bytes reclaimed", That looks strange.
Marcus
Am 05.10.2005 um 14:07 schrieb Marcus Denker:
It could be that someting is wrong with the cleaning... I just did a flush in my newly build work-image and it said: "3 versions flushed, 344bytes reclaimed", That looks strange.
Well, the output is not too accurate - it just does a full GC, remembers the free space, then does a full GC again and reports the difference:
flushAllCaches | beforeBytes afterBytes beforeVersions afterVersions | Cursor wait showWhile: [ beforeBytes := Smalltalk garbageCollect. beforeVersions := MCVersion allSubInstances size. MCFileBasedRepository flushAllCaches. afterBytes := Smalltalk garbageCollect. afterVersions := MCVersion allSubInstances size. ]. ^self inform: (beforeVersions - afterVersions) asString, ' versions flushed', String cr, (afterBytes - beforeBytes) asStringWithCommas, ' bytes reclaimed'
If the total memory was shrunken by the second GC then the report would be way off.
You could try these:
MCVersion allSubInstances size MCDefinition allSubInstances size MCPatchOperation allSubInstances size
which should all report 0 after flushing.
- Bert -
Am 05.10.2005 um 14:41 schrieb Bert Freudenberg:
Am 05.10.2005 um 14:07 schrieb Marcus Denker:
It could be that someting is wrong with the cleaning... I just did a flush in my newly build work-image and it said: "3 versions flushed, 344bytes reclaimed", That looks strange.
You could try these:
MCVersion allSubInstances size MCDefinition allSubInstances size MCPatchOperation allSubInstances size
MCDefinition allSubInstances size : 9604
Marcus
packages@lists.squeakfoundation.org
-
Bert Freudenberg -
Cees De Groot -
Marcus Denker -
stéphane ducasse