[opensuse] 32bit 'tree' doesn't show files 2G and larger - packaging error?
Hi Before submitting to bugzilla, I'd like to get some second opinions... Problem: 32bit openSUSE 11.2; jon@nx8220:~> ll tmp/largefiles/ total 4194304 -rw-r--r-- 1 jon users 2147483648 2010-04-30 12:50 2G -rw-r--r-- 1 jon users 2147482624 2010-04-30 12:52 2G-1k jon@nx8220:~> tree tmp/largefiles/ tmp/largefiles/ `-- 2G-1k 0 directories, 1 file 64bit openSUSE 11.2; jon@cnu8092slm:~> ll tmp/largefiles/ total 4194304 -rw-r--r-- 1 jon users 2147483648 2010-04-30 12:58 2G -rw-r--r-- 1 jon users 2147482624 2010-04-30 12:58 2G-1k jon@cnu8092slm:~> tree tmp/largefiles/ tmp/largefiles/ |-- 2G `-- 2G-1k 0 directories, 2 files On both systems the files were created with: dd if=/dev/zero of=2G bs=1k count=2097152 dd if=/dev/zero of=2G-1k bs=1k count=2097151 After chasing around a bit (comparing compiler flags on openSUSE vs. Debian, where 'tree' works on 32bit) I *think* I've found the problem: In tree.spec the 'make' command is: %build make CFLAGS="$RPM_OPT_FLAGS" -which overrides the CFLAGS set in the Makefile: CFLAGS=-O2 -Wall -fomit-frame-pointer -DLINUX -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -which results in the binary being built with these flags instead: RPM_OPT_FLAGS="-O2 -g -m32 -march=i586 -mtune=i686 -fmessage-length=0 -D_FORTIFY_SOURCE=2 -fstack-protector -funwind-tables -fasynchronous-unwind-tables" Now; If I change the spec file so that; %build make -then the compiler gets the flags as set in the Makefile, and the resulting binary works with files larger than 2G. Question: I'm guessing that it's desirable to get $RPM_OPT_FLAGS passed to the compiler, so how does one 'merge' this variable with $CFLAGS from the Makefile? AFAICT the flag that makes the difference is -D_FILE_OFFSET_BITS=64 - which makes the binary use *stat64() rather than *stat() when files are 'large'. On 64bit systems I suppose everything uses *stat64() anyways, which is why the problem only surfaces on 32bit systems... ? TIA /jon -- YMMV -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Sunday, 2010-05-02 at 14:02 +0200, Jon Clausen wrote:
Before submitting to bugzilla, I'd like to get some second opinions...
I can reproduce on my 11.0, i586 system. My 11.2 on another computer is x64. I have another 11.2 i586 on that computer, but can't boot it right now, busy machine. I could try 11.3 on the laptop, perhaps... cer@nimrodel:~> l /data/Grande/cer/distros/ -lh total 7.1G drwxr-xr-x 3 cer users 4.0K 2009-11-27 15:37 ./ drwxr-xr-x 13 cer root 4.0K 2009-12-17 01:25 ../ - -rw-r--r-- 1 cer users 136 2008-10-22 09:59 MD5SUMS.111 - -rw-r--r-- 1 cer users 21K 2008-11-21 18:00 SGD_Howto_make.html - -rw-r--r-- 1 cer users 653M 2008-04-21 16:43 andlinux-beta1-kde.exe - -rw-r--r-- 1 cer users 747 2008-01-26 13:34 lliurex.tar.gz - -rw-r--r-- 1 cer users 45 2009-02-09 19:50 notas drwxr-xr-x 2 cer users 131 2009-07-19 15:00 openSUSE-11.1-DVD-i586.iso/ - -rw-r--r-- 1 cer users 11K 2008-12-19 20:57 openSUSE-11.1-DVD-i586.iso.metalink - -rw-r--r-- 1 cer users 4.2G 2009-11-15 00:12 openSUSE-11.2-DVD-i586.iso - -rw-r--r-- 1 cer users 23K 2009-11-14 11:57 openSUSE-11.2-DVD-i586.iso.metalink cer@nimrodel:~> tree /data/Grande/cer/distros/ /data/Grande/cer/distros/ |-- MD5SUMS.111 |-- SGD_Howto_make.html |-- andlinux-beta1-kde.exe |-- lliurex.tar.gz |-- notas |-- openSUSE-11.1-DVD-i586.iso | |-- openSUSE-11.1-DVD-i586.iso.asc | |-- openSUSE-11.1-DVD-i586.iso.md5 | `-- openSUSE-11.1-DVD-i586.iso__borrado_20090719 |-- openSUSE-11.1-DVD-i586.iso.metalink |-- openSUSE-11.2-DVD-i586.iso.metalink The +4 GB 11.2 iso image is missing. Go ahead, report. IMHO. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.9 (GNU/Linux) iEYEARECAAYFAkvdcQsACgkQtTMYHG2NR9W2SgCdFOH2cJen9LOy2V7RTvxanvY6 bI4AoIayEtWHEScY7XQh+Xvv7Wc78SX+ =1Okj -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Sun, 02 May, 2010 at 14:33:07 +0200, Carlos E. R. wrote:
On Sunday, 2010-05-02 at 14:02 +0200, Jon Clausen wrote:
Before submitting to bugzilla, I'd like to get some second opinions...
I can reproduce on my 11.0, i586 system. My 11.2 on another computer is x64. I have another 11.2 i586 on that computer, but can't boot it right now, busy machine. I could try 11.3 on the laptop, perhaps...
Thanks :) <snip>
Go ahead, report. IMHO.
Will do, but I'll wait a while so someone else has a chance to respond to the 'merge' Makefile vs. RPM_OPT_FLAGS variables question... /jon -- YMMV -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Sun, 02 May, 2010 at 14:48:34 +0200, Jon Clausen wrote:
On Sun, 02 May, 2010 at 14:33:07 +0200, Carlos E. R. wrote:
Go ahead, report. IMHO.
Will do, but I'll wait a while so someone else has a chance to respond to the 'merge' Makefile vs. RPM_OPT_FLAGS variables question...
Hmmm... On 2010-05-02 I created https://bugzilla.novell.com/show_bug.cgi?id=601779 On 2010-05-04 Marcus wrote: https://bugzilla.novell.com/show_bug.cgi?id=601779#c1 Can someone please help me understand what this means? Is it something about the priority/severity of the bug that makes the fixed package not show up as an update? I mean it seems pretty clear to me, that * the package is broken as shipped with the distribution * the fixed package should show up as a regular update so why doesn't it? tia /jon -- YMMV -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On 12/05/10 11:10, Jon Clausen wrote:
On 2010-05-02 I created https://bugzilla.novell.com/show_bug.cgi?id=601779 On 2010-05-04 Marcus wrote: https://bugzilla.novell.com/show_bug.cgi?id=601779#c1
Can someone please help me understand what this means?
Is it something about the priority/severity of the bug that makes the fixed package not show up as an update?
I mean it seems pretty clear to me, that * the package is broken as shipped with the distribution * the fixed package should show up as a regular update
so why doesn't it?
To be released as an online update for 11.2 or previous, first it has to be requested by the people involved then approved by the maintenance team then pushed into the SUSE internal update system. If you feel this bug is important enough to warrant an online update, make your case for it in the bug. After this, it can still take some weeks to enter into the update system, and then be pushed out to users. Have a look at http://en.opensuse.org/Maintenance https://build.opensuse.org/maintenance It looks like the fix for the bug you mention has gone into a devel project, which is only step one for even getting into Factory. So far noone seems to have requested it be pushed into previous releases. Just my opinion, but I think that the case for this being an update is a bit poor, because you say it has been broken for a long time and noone seems to have noticed till now. But on the other hand, it would be a very simple update - only a single package is affected. Regards, Tejas -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Wed, May 12, 2010 at 11:36:12AM +0100, Tejas Guruswamy wrote:
On 12/05/10 11:10, Jon Clausen wrote:
On 2010-05-02 I created https://bugzilla.novell.com/show_bug.cgi?id=601779 On 2010-05-04 Marcus wrote: https://bugzilla.novell.com/show_bug.cgi?id=601779#c1
Can someone please help me understand what this means?
Is it something about the priority/severity of the bug that makes the fixed package not show up as an update?
I mean it seems pretty clear to me, that * the package is broken as shipped with the distribution * the fixed package should show up as a regular update
so why doesn't it?
To be released as an online update for 11.2 or previous, first it has to be requested by the people involved then approved by the maintenance team then pushed into the SUSE internal update system.
If you feel this bug is important enough to warrant an online update, make your case for it in the bug. After this, it can still take some weeks to enter into the update system, and then be pushed out to users.
Have a look at http://en.opensuse.org/Maintenance https://build.opensuse.org/maintenance
It looks like the fix for the bug you mention has gone into a devel project, which is only step one for even getting into Factory. So far noone seems to have requested it be pushed into previous releases.
Just my opinion, but I think that the case for this being an update is a bit poor, because you say it has been broken for a long time and noone seems to have noticed till now. But on the other hand, it would be a very simple update - only a single package is affected.
The fix is quite simple and I did it only for Factory... In general we do not fix everything for older distributions ... For this specific bug I wonder if the usecase of "tree" displaying files larger than 2GB is really a bug with high impact that we should spend time on. I can reopen and NEEDINFO maintenance@opensuse.org ... Ciao, Marcus -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Wed, 12 May, 2010 at 13:18:25 +0200, Marcus Meissner wrote:
On Wed, May 12, 2010 at 11:36:12AM +0100, Tejas Guruswamy wrote:
On 12/05/10 11:10, Jon Clausen wrote:
I mean it seems pretty clear to me, that * the package is broken as shipped with the distribution * the fixed package should show up as a regular update
so why doesn't it?
To be released as an online update for 11.2 or previous, first it has to be requested by the people involved then approved by the maintenance team then pushed into the SUSE internal update system.
Hmm... a lot more involved than I thought. <snip>
Just my opinion, but I think that the case for this being an update is a bit poor, because you say it has been broken for a long time and noone seems to have noticed till now.
Well, maybe I'm just the first to be sufficiently annoyed to report the bug?
But on the other hand, it would be a very simple update - only a single package is affected.
Precisely; The fix is trivial, only one package is affected, so why *not* push out as an update?
The fix is quite simple and I did it only for Factory...
In general we do not fix everything for older distributions ...
Of course I realize that stuff has to go through a screening process, testing etc - which takes resources - before it can be pushed out as an Update... but; It makes me wonder how many other faulty packages I have on my systems, where commands have slight malfunctions, which I may overlook? Don't get me wrong; I do appreciate the quick fix. But I am a little surprised that *any* "it's clearly a bug, which is easily fixed" doesn't just get submitted to the Updates 'tree', rather than having to pass a "usecase evaluation threshold"...
For this specific bug I wonder if the usecase of "tree" displaying files larger than 2GB is really a bug with high impact that we should spend time on.
Well for me at least, it has a pretty high impact in that several of my filesystems have many files way larger than that.
I can reopen and NEEDINFO maintenance@opensuse.org ...
Yes, please :) Should I do something? thanks /jon -- YMMV -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Wednesday, 2010-05-12 at 14:07 +0200, Jon Clausen wrote: ...
It makes me wonder how many other faulty packages I have on my systems, where commands have slight malfunctions, which I may overlook?
Probably a lot. Many get "repaired", but the repair is often only available on the next system release - which, of course, will contain its own bucketfulls of newer bugs. Is how software gets done. The important point is new features, not getting bugless software. Instead of stopping and clearing all bugs, we get lots of new features, with some old bugs solved, and new bugs added. >:-)
Should I do something?
Well, if it is a problem for you (tree), perhaps you can add the patch and recompile yourself, or perhaps some kind soul does it somewhere in the buildservice :-) It is not a problem for me because I do not use tree much. I think... - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.9 (GNU/Linux) iEYEARECAAYFAkvrQQoACgkQtTMYHG2NR9XL4wCdG1A4EZ/+PPZmL9lQfG5t2VA9 UDQAnR1042xtaUOVVxm7aeSFokDJOPw4 =mg+L -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Thu, 13 May, 2010 at 02:00:03 +0200, Carlos E. R. wrote:
On Wednesday, 2010-05-12 at 14:07 +0200, Jon Clausen wrote:
...
It makes me wonder how many other faulty packages I have on my systems, where commands have slight malfunctions, which I may overlook?
Probably a lot.
...please don't say that :P
Many get "repaired", but the repair is often only available on the next system release - which, of course, will contain its own bucketfulls of newer bugs.
Probably so. And to some extent 'fair enough', that fixes don't automatically get 'backported'. But I'm still somewhat disappointed that one has 'plead the case' in order for a fix, to a clearly documented bug, to be 'promoted' to an Update.
Is how software gets done. The important point is new features, not getting bugless software. Instead of stopping and clearing all bugs, we get lots of new features, with some old bugs solved, and new bugs added. >:-)
Yeah, I guess you're right. I wasn't entirely clear in what I wrote in the above quote. What I meant was not so much 'packages built from faulty source code', but rather 'packages that are known to be faultily built, from (maybe faulty) source code'. What I think I'm trying to say is; If it's worth packaging, then it's worth packaging *correctly*. But also; IMO packaging errors shouldn't be subject to a 'popularity contest'. Scrutiny and validation... of course, but the quality of the distribution is in large part 'quality of packaging'... isn't it?
Should I do something?
Well, if it is a problem for you (tree), perhaps you can add the patch and recompile yourself, or perhaps some kind soul does it somewhere in the buildservice :-)
Yeah well, I think I'll abstain from attempting to recomplie *myself* ;) What I meant was more in the line of "should I do something, add comments or whatever, when the bug is reopened?"
It is not a problem for me because I do not use tree much. I think...
I use tree enough that it's certainly worth rebuilding the rpm for myself. With the research I put into tracking down the issue in the first place, it's not even going to be that big of a deal - especially now that I've found the src.rpm Marcus fixed - so I can see *how* he fixed it. But I'd much rather just 'zypper up', and get the fixed package that way :P cheers /jon -- YMMV -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Thu, May 13, 2010 at 09:43:49AM +0200, Jon Clausen wrote:
On Thu, 13 May, 2010 at 02:00:03 +0200, Carlos E. R. wrote:
On Wednesday, 2010-05-12 at 14:07 +0200, Jon Clausen wrote:
...
It makes me wonder how many other faulty packages I have on my systems, where commands have slight malfunctions, which I may overlook?
Probably a lot.
...please don't say that :P
Many get "repaired", but the repair is often only available on the next system release - which, of course, will contain its own bucketfulls of newer bugs.
Probably so. And to some extent 'fair enough', that fixes don't automatically get 'backported'. But I'm still somewhat disappointed that one has 'plead the case' in order for a fix, to a clearly documented bug, to be 'promoted' to an Update.
Is how software gets done. The important point is new features, not getting bugless software. Instead of stopping and clearing all bugs, we get lots of new features, with some old bugs solved, and new bugs added. >:-)
Yeah, I guess you're right. I wasn't entirely clear in what I wrote in the above quote. What I meant was not so much 'packages built from faulty source code', but rather 'packages that are known to be faultily built, from (maybe faulty) source code'.
What I think I'm trying to say is; If it's worth packaging, then it's worth packaging *correctly*. But also;
IMO packaging errors shouldn't be subject to a 'popularity contest'. Scrutiny and validation... of course, but the quality of the distribution is in large part 'quality of packaging'... isn't it?
Should I do something?
Well, if it is a problem for you (tree), perhaps you can add the patch and recompile yourself, or perhaps some kind soul does it somewhere in the buildservice :-)
Yeah well, I think I'll abstain from attempting to recomplie *myself* ;)
What I meant was more in the line of "should I do something, add comments or whatever, when the bug is reopened?"
It is not a problem for me because I do not use tree much. I think...
I use tree enough that it's certainly worth rebuilding the rpm for myself. With the research I put into tracking down the issue in the first place, it's not even going to be that big of a deal - especially now that I've found the src.rpm Marcus fixed - so I can see *how* he fixed it.
But I'd much rather just 'zypper up', and get the fixed package that way :P
The bug was approved to be fixed for 11.2, so I will prepare an update. Ciao, Marcus -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Sun, 16 May, 2010 at 14:44:41 +0200, Marcus Meissner wrote:
On Thu, May 13, 2010 at 09:43:49AM +0200, Jon Clausen wrote:
But I'd much rather just 'zypper up', and get the fixed package that way :P
The bug was approved to be fixed for 11.2, so I will prepare an update.
Thanks :) /jon -- YMMV -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
El 12/05/10 07:18, Marcus Meissner escribió:
The fix is quite simple and I did it only for Factory...
Yep, but the inner problem is still there, there is no check in the buildsystem to detect this type of issues. Not to mention that releasing LFS updates might break the ABI... :) -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
participants (5)
-
Carlos E. R. -
Cristian Rodríguez -
Jon Clausen -
Marcus Meissner -
Tejas Guruswamy