Huge patch
|
View:
New views
3 Messages
—
Rating Filter:
Alert me
|
 |
Huge patch
by Rob Braun
::
Rate this Message:
1) I disabled macho parsing all together. This is temporary.
2) There was a bug that presumably predated the patch, but making the compression state modular made the bug critical. This was using the context after it was deallocated in *_fromheap_done() in both the gzip and bzip2 compression modules. I encountered some problems while extracting hardlinks, which should probably be looked into. Overall, thanks Chris, this is a huge improvement in xar's abilities! Thanks, Rob On Wed, Jul 19, 2006 at 09:39:34AM -0700, Christopher Ryan wrote: > I will try making large archives (with large number of files) to see > what happens. The mach-o changes in my diff actually back-rev the > mach-o changes in tot which makes a buffer for each file it reads in, > so it should be more stable. > > I am surprised you can even make an archive with a toc that doesn't > verify considering the checksum of the toc is calculated as the toc is > put into the actual file. > > Let me know what you find in regards to the mach-o part. > > Chris > > > >Ok I finally got around to doing this. Looks like we might have to > >totally disable mach-o parsing until it's correct. I tried to dump > >the toc header from a 24GB xar archive containing Christopher's > >changes. It didn't pass a checksum test. > > > >I'll disable mach-o and re-run tonight. > > > >I also verified that the xar I had before, without mach-o, was working > >fine with --dump-toc. > > > >Dave > > > >On 7/16/06, David Leimbach <leimy2k@...> wrote: > >>Just to let you know I've tried out your patches on a few things this > >>evening, and so far so good as far as I'm concerned. I typically do > >>very large backups at night too with xar and make archives that are > >>on > >>the order of 10-15 GB. > >> > >>I'm giving it a shot tonight with the patches you've sent applied. > >> > >>Dave > >> > >>On 7/12/06, Christopher Ryan <cnr1089@...> wrote: > >>> Ok, here are the changes I have been talking about. Please let me > >>> know what you guys think. Feedback is appreciated. I tried to > >>stick > >>> to the coding conventions already in xar as much as possible. I > >>> haven't had a chance to compile this on any other platform other > >>then > >>> Mac OS X (so let me know if there are problems, although I don't > >>> expect any), I apologize for this. > >>> > >>> 1) Makes the xar.dylib more flexible as a library by changing many > >>> global variables into context data. > >>> 2) Added support for different heap data hash algorithms (none, > >>md5, > >>> or sha1) > >>> 3) Added support for working with xar files in memory (rather > >>then a > >>> file on disk) > >>> 4) Added a validation method which validates a section of the heap > >>> against the checksum (if any) without having to extract to disk > >>> 5) Added support for the "signature" section in the toc as > >>proposed by > >>> Kevin a couple of months back. > >>> This does not do signing, just supports a client who may want to > >>> sign. As of right now it only has hooks for storing X509 Certs. > >>> Additional hooks can be added. > >>> > >>> Additionally, as I have noted in my last few e-mails, I have: > >>> > >>> 1) revered the mach-o changes that went in between 1.4 and TOT. > >>> > >>> > >>> Thanks, > >>> > >>> Chris > >>> > >>> > >>> > >>> > >>> > >>> > >>> > >>> > >>> _______________________________________________ > >>> xar mailing list > >>> xar@... > >>> http://www.opendarwin.org/mailman/listinfo/xar > >>> > >>> > >>> > >>> > >> > > _______________________________________________ > xar mailing list > xar@... > http://www.opendarwin.org/mailman/listinfo/xar _______________________________________________ xar mailing list xar@... http://www.opendarwin.org/mailman/listinfo/xar |
|
|
Re: Huge patch
by David Leimbach
::
Rate this Message:
Anyone testing xar as a 64bit library/binary anywhere?
I've got one I need a reason to power on I think... maybe when the temperature drops though... it runs a wee hot. Dave On 7/19/06, Rob Braun <bbraun@...> wrote: > 1) I disabled macho parsing all together. This is temporary. > 2) There was a bug that presumably predated the patch, but > making the compression state modular made the bug critical. > This was using the context after it was deallocated in *_fromheap_done() > in both the gzip and bzip2 compression modules. > > I encountered some problems while extracting hardlinks, which should > probably be looked into. > > Overall, thanks Chris, this is a huge improvement in xar's abilities! > > Thanks, > Rob > > On Wed, Jul 19, 2006 at 09:39:34AM -0700, Christopher Ryan wrote: > > I will try making large archives (with large number of files) to see > > what happens. The mach-o changes in my diff actually back-rev the > > mach-o changes in tot which makes a buffer for each file it reads in, > > so it should be more stable. > > > > I am surprised you can even make an archive with a toc that doesn't > > verify considering the checksum of the toc is calculated as the toc is > > put into the actual file. > > > > Let me know what you find in regards to the mach-o part. > > > > Chris > > > > > > >Ok I finally got around to doing this. Looks like we might have to > > >totally disable mach-o parsing until it's correct. I tried to dump > > >the toc header from a 24GB xar archive containing Christopher's > > >changes. It didn't pass a checksum test. > > > > > >I'll disable mach-o and re-run tonight. > > > > > >I also verified that the xar I had before, without mach-o, was working > > >fine with --dump-toc. > > > > > >Dave > > > > > >On 7/16/06, David Leimbach <leimy2k@...> wrote: > > >>Just to let you know I've tried out your patches on a few things this > > >>evening, and so far so good as far as I'm concerned. I typically do > > >>very large backups at night too with xar and make archives that are > > >>on > > >>the order of 10-15 GB. > > >> > > >>I'm giving it a shot tonight with the patches you've sent applied. > > >> > > >>Dave > > >> > > >>On 7/12/06, Christopher Ryan <cnr1089@...> wrote: > > >>> Ok, here are the changes I have been talking about. Please let me > > >>> know what you guys think. Feedback is appreciated. I tried to > > >>stick > > >>> to the coding conventions already in xar as much as possible. I > > >>> haven't had a chance to compile this on any other platform other > > >>then > > >>> Mac OS X (so let me know if there are problems, although I don't > > >>> expect any), I apologize for this. > > >>> > > >>> 1) Makes the xar.dylib more flexible as a library by changing many > > >>> global variables into context data. > > >>> 2) Added support for different heap data hash algorithms (none, > > >>md5, > > >>> or sha1) > > >>> 3) Added support for working with xar files in memory (rather > > >>then a > > >>> file on disk) > > >>> 4) Added a validation method which validates a section of the heap > > >>> against the checksum (if any) without having to extract to disk > > >>> 5) Added support for the "signature" section in the toc as > > >>proposed by > > >>> Kevin a couple of months back. > > >>> This does not do signing, just supports a client who may want to > > >>> sign. As of right now it only has hooks for storing X509 Certs. > > >>> Additional hooks can be added. > > >>> > > >>> Additionally, as I have noted in my last few e-mails, I have: > > >>> > > >>> 1) revered the mach-o changes that went in between 1.4 and TOT. > > >>> > > >>> > > >>> Thanks, > > >>> > > >>> Chris > > >>> > > >>> > > >>> > > >>> > > >>> > > >>> > > >>> > > >>> > > >>> _______________________________________________ > > >>> xar mailing list > > >>> xar@... > > >>> http://www.opendarwin.org/mailman/listinfo/xar > > >>> > > >>> > > >>> > > >>> > > >> > > > > _______________________________________________ > > xar mailing list > > xar@... > > http://www.opendarwin.org/mailman/listinfo/xar > > _______________________________________________ > xar mailing list > xar@... > http://www.opendarwin.org/mailman/listinfo/xar > xar mailing list xar@... http://www.opendarwin.org/mailman/listinfo/xar |
|
|
|
| Free embeddable forum powered by Nabble | Forum Help |
