Re: [Vm-dev] Bug 956376 ??? Can not edit class comments
Locked
Re: [Vm-dev] Bug 956376 ??? Can not edit class comments
 
|
On 25 April 2013 15:35, Frank Shearar <[hidden email]> wrote:
> On 25 April 2013 13:54, David T. Lewis <[hidden email]> wrote: >> >> On Thu, Apr 25, 2013 at 02:18:22PM +0200, Bert Freudenberg wrote: >>> >>> This didn't sound like a VM problem at first, but according to the discussion it is: >>> >>> https://bugzilla.redhat.com/show_bug.cgi?id=956376 >>> >>> Any idea? >>> >> >> On my 64-bit Linux with an interpreter VM compiled in 64-bit, and using >> Squeak - no problem. With a 32-bit interpreter - no problem. With Cog - >> again, no problem. >> >> I edited the class comment for class Array in each case. >> >> Dave > > "Error: RemoteString past end of file" can indicate a broken/damages > changes file. > > It does when I get the error, at least. Usually that happens when I > have to hard-kill an image that's stored state in the changes file. > Restarting the image means there's a mismatch somehow in the state > between the image and the changes file. Once you get one of these > you're nearly completely hosed: you can't file things out, for > instance. For those not following the Pharo list, it looks like the root cause is that the glibc in Ubuntu 13.04 doesn't immediately write new content out to the file. Thus you can "write" to the file, and then read from the file, only your RemoteString starts off the end of the file. The fix is to make WriteStream >> #nextPutChunk: call "self flush". I've pushed a 4.5 fix to trunk, but have not yet done the same for the 4.4 and 4.3 update streams. frank |
Re: [Vm-dev] Bug 956376 ??? Can not edit class comments
|
|
Please submit then info to redhat bugzilla about where Jaroslav can find the fixed 4.3 image to repack it.
-- Pavel
On Fri, Apr 26, 2013 at 12:17 PM, Frank Shearar <[hidden email]> wrote:
|
Re: [Vm-dev] Bug 956376 ??? Can not edit class comments
|
|
In reply to this post by Frank Shearar-3
On Fri, Apr 26, 2013 at 11:17:04AM +0100, Frank Shearar wrote:
> On 25 April 2013 15:35, Frank Shearar <[hidden email]> wrote: > > On 25 April 2013 13:54, David T. Lewis <[hidden email]> wrote: > >> > >> On Thu, Apr 25, 2013 at 02:18:22PM +0200, Bert Freudenberg wrote: > >>> > >>> This didn't sound like a VM problem at first, but according to the discussion it is: > >>> > >>> https://bugzilla.redhat.com/show_bug.cgi?id=956376 > >>> > >>> Any idea? > >>> > >> > >> On my 64-bit Linux with an interpreter VM compiled in 64-bit, and using > >> Squeak - no problem. With a 32-bit interpreter - no problem. With Cog - > >> again, no problem. > >> > >> I edited the class comment for class Array in each case. > >> > >> Dave > > > > "Error: RemoteString past end of file" can indicate a broken/damages > > changes file. > > > > It does when I get the error, at least. Usually that happens when I > > have to hard-kill an image that's stored state in the changes file. > > Restarting the image means there's a mismatch somehow in the state > > between the image and the changes file. Once you get one of these > > you're nearly completely hosed: you can't file things out, for > > instance. > > For those not following the Pharo list, it looks like the root cause > is that the glibc in Ubuntu 13.04 doesn't immediately write new > content out to the file. Thus you can "write" to the file, and then > read from the file, only your RemoteString starts off the end of the > file. The fix is to make WriteStream >> #nextPutChunk: call "self > flush". I've pushed a 4.5 fix to trunk, but have not yet done the same > for the 4.4 and 4.3 update streams. > How about Magma? If writing to positionable files is broken, then it's going to affect a lot more than just writing to the .changes file. Has this bug been observed with any of the VMs distributed for Squeak? I cannot tell from the discussion on the Pharo list, and I don't have access to a Ubuntu 13.04 system to try it for myself. Dave |
Re: [Vm-dev] Bug 956376 ??? Can not edit class comments
|
|
In reply to this post by Pavel Krivanek
Agreed. It's more complicated than it sounds, of course: the proper
fix is to push out new updated binaries to ftp.squeak.org. frank On 26 April 2013 13:17, Pavel Krivanek <[hidden email]> wrote: > Please submit then info to redhat bugzilla about where Jaroslav can find the > fixed 4.3 image to repack it. > > -- Pavel > > > On Fri, Apr 26, 2013 at 12:17 PM, Frank Shearar <[hidden email]> > wrote: >> >> On 25 April 2013 15:35, Frank Shearar <[hidden email]> wrote: >> > On 25 April 2013 13:54, David T. Lewis <[hidden email]> wrote: >> >> >> >> On Thu, Apr 25, 2013 at 02:18:22PM +0200, Bert Freudenberg wrote: >> >>> >> >>> This didn't sound like a VM problem at first, but according to the >> >>> discussion it is: >> >>> >> >>> https://bugzilla.redhat.com/show_bug.cgi?id=956376 >> >>> >> >>> Any idea? >> >>> >> >> >> >> On my 64-bit Linux with an interpreter VM compiled in 64-bit, and using >> >> Squeak - no problem. With a 32-bit interpreter - no problem. With Cog - >> >> again, no problem. >> >> >> >> I edited the class comment for class Array in each case. >> >> >> >> Dave >> > >> > "Error: RemoteString past end of file" can indicate a broken/damages >> > changes file. >> > >> > It does when I get the error, at least. Usually that happens when I >> > have to hard-kill an image that's stored state in the changes file. >> > Restarting the image means there's a mismatch somehow in the state >> > between the image and the changes file. Once you get one of these >> > you're nearly completely hosed: you can't file things out, for >> > instance. >> >> For those not following the Pharo list, it looks like the root cause >> is that the glibc in Ubuntu 13.04 doesn't immediately write new >> content out to the file. Thus you can "write" to the file, and then >> read from the file, only your RemoteString starts off the end of the >> file. The fix is to make WriteStream >> #nextPutChunk: call "self >> flush". I've pushed a 4.5 fix to trunk, but have not yet done the same >> for the 4.4 and 4.3 update streams. >> >> frank >> > > > > |
Re: [Vm-dev] Bug 956376 ??? Can not edit class comments
|
|
Frank,
Can you please clarify if this is a bug that has been observed with any of the *existing* pre-compiled VMs obtained from either http://squeakvm.org/unix/ or from http://www.mirandabanda.org/files/Cog/VM/? Adding a workaround to one spot in the image may be the wrong thing to do. It would mask a serious problem that could lead to data corruption for anyone who relies on storing data in files. I am traveling and will not be able to follow up on this mail for a while, but please - if this is a gcc optimizer problem, as has been suggested elsewhere in the discussion, then let's not "fix" it in the image. Dave On Fri, Apr 26, 2013 at 01:52:55PM +0100, Frank Shearar wrote: > Agreed. It's more complicated than it sounds, of course: the proper > fix is to push out new updated binaries to ftp.squeak.org. > > frank > > On 26 April 2013 13:17, Pavel Krivanek <[hidden email]> wrote: > > Please submit then info to redhat bugzilla about where Jaroslav can find the > > fixed 4.3 image to repack it. > > > > -- Pavel > > > > > > On Fri, Apr 26, 2013 at 12:17 PM, Frank Shearar <[hidden email]> > > wrote: > >> > >> On 25 April 2013 15:35, Frank Shearar <[hidden email]> wrote: > >> > On 25 April 2013 13:54, David T. Lewis <[hidden email]> wrote: > >> >> > >> >> On Thu, Apr 25, 2013 at 02:18:22PM +0200, Bert Freudenberg wrote: > >> >>> > >> >>> This didn't sound like a VM problem at first, but according to the > >> >>> discussion it is: > >> >>> > >> >>> https://bugzilla.redhat.com/show_bug.cgi?id=956376 > >> >>> > >> >>> Any idea? > >> >>> > >> >> > >> >> On my 64-bit Linux with an interpreter VM compiled in 64-bit, and using > >> >> Squeak - no problem. With a 32-bit interpreter - no problem. With Cog - > >> >> again, no problem. > >> >> > >> >> I edited the class comment for class Array in each case. > >> >> > >> >> Dave > >> > > >> > "Error: RemoteString past end of file" can indicate a broken/damages > >> > changes file. > >> > > >> > It does when I get the error, at least. Usually that happens when I > >> > have to hard-kill an image that's stored state in the changes file. > >> > Restarting the image means there's a mismatch somehow in the state > >> > between the image and the changes file. Once you get one of these > >> > you're nearly completely hosed: you can't file things out, for > >> > instance. > >> > >> For those not following the Pharo list, it looks like the root cause > >> is that the glibc in Ubuntu 13.04 doesn't immediately write new > >> content out to the file. Thus you can "write" to the file, and then > >> read from the file, only your RemoteString starts off the end of the > >> file. The fix is to make WriteStream >> #nextPutChunk: call "self > >> flush". I've pushed a 4.5 fix to trunk, but have not yet done the same > >> for the 4.4 and 4.3 update streams. > >> > >> frank > >> > > > > > > > > |
Re: [Vm-dev] Bug 956376 ??? Can not edit class comments
|
|
I have not seen this error in any of Ubuntu 10.10, Windows, OS X
(whatever version I have - 10.4? 10.6?) or CentOS, with any of the Cog, CogMT, Interpreter VMs. But I _don't_ have an Ubuntu 13.04 machine. frank On 26 April 2013 14:55, David T. Lewis <[hidden email]> wrote: > Frank, > > Can you please clarify if this is a bug that has been observed with any of > the *existing* pre-compiled VMs obtained from either http://squeakvm.org/unix/ > or from http://www.mirandabanda.org/files/Cog/VM/? > > Adding a workaround to one spot in the image may be the wrong thing to do. > It would mask a serious problem that could lead to data corruption for > anyone who relies on storing data in files. > > I am traveling and will not be able to follow up on this mail for a while, > but please - if this is a gcc optimizer problem, as has been suggested > elsewhere in the discussion, then let's not "fix" it in the image. > > Dave > > > On Fri, Apr 26, 2013 at 01:52:55PM +0100, Frank Shearar wrote: >> Agreed. It's more complicated than it sounds, of course: the proper >> fix is to push out new updated binaries to ftp.squeak.org. >> >> frank >> >> On 26 April 2013 13:17, Pavel Krivanek <[hidden email]> wrote: >> > Please submit then info to redhat bugzilla about where Jaroslav can find the >> > fixed 4.3 image to repack it. >> > >> > -- Pavel >> > >> > >> > On Fri, Apr 26, 2013 at 12:17 PM, Frank Shearar <[hidden email]> >> > wrote: >> >> >> >> On 25 April 2013 15:35, Frank Shearar <[hidden email]> wrote: >> >> > On 25 April 2013 13:54, David T. Lewis <[hidden email]> wrote: >> >> >> >> >> >> On Thu, Apr 25, 2013 at 02:18:22PM +0200, Bert Freudenberg wrote: >> >> >>> >> >> >>> This didn't sound like a VM problem at first, but according to the >> >> >>> discussion it is: >> >> >>> >> >> >>> https://bugzilla.redhat.com/show_bug.cgi?id=956376 >> >> >>> >> >> >>> Any idea? >> >> >>> >> >> >> >> >> >> On my 64-bit Linux with an interpreter VM compiled in 64-bit, and using >> >> >> Squeak - no problem. With a 32-bit interpreter - no problem. With Cog - >> >> >> again, no problem. >> >> >> >> >> >> I edited the class comment for class Array in each case. >> >> >> >> >> >> Dave >> >> > >> >> > "Error: RemoteString past end of file" can indicate a broken/damages >> >> > changes file. >> >> > >> >> > It does when I get the error, at least. Usually that happens when I >> >> > have to hard-kill an image that's stored state in the changes file. >> >> > Restarting the image means there's a mismatch somehow in the state >> >> > between the image and the changes file. Once you get one of these >> >> > you're nearly completely hosed: you can't file things out, for >> >> > instance. >> >> >> >> For those not following the Pharo list, it looks like the root cause >> >> is that the glibc in Ubuntu 13.04 doesn't immediately write new >> >> content out to the file. Thus you can "write" to the file, and then >> >> read from the file, only your RemoteString starts off the end of the >> >> file. The fix is to make WriteStream >> #nextPutChunk: call "self >> >> flush". I've pushed a 4.5 fix to trunk, but have not yet done the same >> >> for the 4.4 and 4.3 update streams. >> >> >> >> frank >> >> >> > >> > >> > >> > > |
Re: [Vm-dev] Bug 956376 ??? Can not edit class comments
|
|
In reply to this post by Frank Shearar-3
Frank are you saying that, under Ubuntu 13.04, after writing to a
particular place in a file, if you subsequently go back and _read_ from that place, even using the same ReadWriteStream instance, that it will not read back in what was just written until a #flush occurs? Hopefully that is not the case! On Fri, Apr 26, 2013 at 5:17 AM, Frank Shearar <[hidden email]> wrote: > On 25 April 2013 15:35, Frank Shearar <[hidden email]> wrote: >> On 25 April 2013 13:54, David T. Lewis <[hidden email]> wrote: >>> >>> On Thu, Apr 25, 2013 at 02:18:22PM +0200, Bert Freudenberg wrote: >>>> >>>> This didn't sound like a VM problem at first, but according to the discussion it is: >>>> >>>> https://bugzilla.redhat.com/show_bug.cgi?id=956376 >>>> >>>> Any idea? >>>> >>> >>> On my 64-bit Linux with an interpreter VM compiled in 64-bit, and using >>> Squeak - no problem. With a 32-bit interpreter - no problem. With Cog - >>> again, no problem. >>> >>> I edited the class comment for class Array in each case. >>> >>> Dave >> >> "Error: RemoteString past end of file" can indicate a broken/damages >> changes file. >> >> It does when I get the error, at least. Usually that happens when I >> have to hard-kill an image that's stored state in the changes file. >> Restarting the image means there's a mismatch somehow in the state >> between the image and the changes file. Once you get one of these >> you're nearly completely hosed: you can't file things out, for >> instance. > > For those not following the Pharo list, it looks like the root cause > is that the glibc in Ubuntu 13.04 doesn't immediately write new > content out to the file. Thus you can "write" to the file, and then > read from the file, only your RemoteString starts off the end of the > file. The fix is to make WriteStream >> #nextPutChunk: call "self > flush". I've pushed a 4.5 fix to trunk, but have not yet done the same > for the 4.4 and 4.3 update streams. > > frank > |
Re: [Vm-dev] Bug 956376 ??? Can not edit class comments
|
|
On Fri, Apr 26, 2013 at 9:25 AM, Chris Muller <[hidden email]> wrote: Frank are you saying that, under Ubuntu 13.04, after writing to a If it is, it is an ubuntu bug and not one we should work around, no?
best, Eliot
|
Re: [Vm-dev] Bug 956376 ??? Can not edit class comments
|
|
On Fri, Apr 26, 2013 at 10:22 AM, Eliot Miranda <[hidden email]> wrote:
If it is a write/read through the same file pointer, that is. If through two separate file pointers then a flush is needed after the write.
best, Eliot
|
Re: [Vm-dev] Bug 956376 ??? Can not edit class comments
|
|
On 26 April 2013 18:23, Eliot Miranda <[hidden email]> wrote:
> > > On Fri, Apr 26, 2013 at 10:22 AM, Eliot Miranda <[hidden email]> > wrote: >> >> >> >> On Fri, Apr 26, 2013 at 9:25 AM, Chris Muller <[hidden email]> wrote: >>> >>> Frank are you saying that, under Ubuntu 13.04, after writing to a >>> particular place in a file, if you subsequently go back and _read_ >>> from that place, even using the same ReadWriteStream instance, that it >>> will not read back in what was just written until a #flush occurs? >>> >>> Hopefully that is not the case! >> >> >> If it is, it is an ubuntu bug and not one we should work around, no? It's highly unlikely to be an _Ubuntu_ bug, especially since there are Fedora versions affected. > If it is a write/read through the same file pointer, that is. If through > two separate file pointers then a flush is needed after the write. Indeed. frank >>> On Fri, Apr 26, 2013 at 5:17 AM, Frank Shearar <[hidden email]> >>> wrote: >>> > On 25 April 2013 15:35, Frank Shearar <[hidden email]> wrote: >>> >> On 25 April 2013 13:54, David T. Lewis <[hidden email]> wrote: >>> >>> >>> >>> On Thu, Apr 25, 2013 at 02:18:22PM +0200, Bert Freudenberg wrote: >>> >>>> >>> >>>> This didn't sound like a VM problem at first, but according to the >>> >>>> discussion it is: >>> >>>> >>> >>>> https://bugzilla.redhat.com/show_bug.cgi?id=956376 >>> >>>> >>> >>>> Any idea? >>> >>>> >>> >>> >>> >>> On my 64-bit Linux with an interpreter VM compiled in 64-bit, and >>> >>> using >>> >>> Squeak - no problem. With a 32-bit interpreter - no problem. With Cog >>> >>> - >>> >>> again, no problem. >>> >>> >>> >>> I edited the class comment for class Array in each case. >>> >>> >>> >>> Dave >>> >> >>> >> "Error: RemoteString past end of file" can indicate a broken/damages >>> >> changes file. >>> >> >>> >> It does when I get the error, at least. Usually that happens when I >>> >> have to hard-kill an image that's stored state in the changes file. >>> >> Restarting the image means there's a mismatch somehow in the state >>> >> between the image and the changes file. Once you get one of these >>> >> you're nearly completely hosed: you can't file things out, for >>> >> instance. >>> > >>> > For those not following the Pharo list, it looks like the root cause >>> > is that the glibc in Ubuntu 13.04 doesn't immediately write new >>> > content out to the file. Thus you can "write" to the file, and then >>> > read from the file, only your RemoteString starts off the end of the >>> > file. The fix is to make WriteStream >> #nextPutChunk: call "self >>> > flush". I've pushed a 4.5 fix to trunk, but have not yet done the same >>> > for the 4.4 and 4.3 update streams. >>> > >>> > frank >>> > >>> >> >> >> >> -- >> best, >> Eliot > > > > > -- > best, > Eliot > > > |
«
Return to Squeak - Dev
|
1 view|%1 views
| Free forum by Nabble | Edit this page |