Subject: No unexpired good images (what should I do?) -
msg#00013

Previous Message by Date:

Spaces in source files - redux

Almost two years ago, one of the very first patches submitted to the dirvish archive was a fix by Eric Mountain that allowed tree names to have spaces in them. Spaces in names are possible with unix, but are much more common with windows. On the wiki, R. Fried pointed out that Eric's fix breaks the regular behavior - it removes the last character of the source tree when there are both sourcetree and aliastree fields . R. Fried suggested removing the patch, which breaks the behavior Eric wants. Here is a patch that hopefully meets both needs, which is described at the end of http://wiki.dirvish.org/index.cgi?SpacesInSource The 1.2.0 behavior by jwschultz looked like so: ($srctree, $aliastree) = split(/\s+/, $$Options{tree}) or seppuku 228, "ERROR: no source tree defined"; $srctree =~ s(/+$)(); $aliastree =~ s(/+$)(); $aliastree ||= $srctree; The 1.2.1 patch by Eric (the current version) looks like: ($srctree, $aliastree) = split(/[^\\]\s+/, $$Options{tree}) or seppuku 228, "ERROR: no source tree defined"; $srctree =~ s(\\ )( )g; $srctree =~ s(/+$)(); $aliastree =~ s(/+$)(); $aliastree ||= $srctree; My suggested patched behavior looks like: $_ = $$Options{tree} ; s/(\s)/\\$1/g; # These two lines swap the escaped spaces with s/\\\\//g; # the nonescaped ones. s/(\\\s)+/$1/g; # replace multiple separators with only one ($srctree,$aliastree) = split /\\\s/ or seppuku 228, "ERROR: no source tree defined"; $aliastree ||= $srctree; For you non-Perl programmers ( which includes me 29 days per month ) the new behavior manipulates the spaces before the split, turning escaped spaces into simple spaces, and simple spaces into escaped spaces. Duplicate escaped spaces are removed, and *then* we do the split. Quite a cheesy hack, but it allows both the original "source and alias" behavior to work, and also the escaped whitespace behavior that Eric wanted to work. In addition, it allows alternate white space characters (like tab) to be preserved in weird directory names, and also allows escaped whitespace on the beginning or end of directory names. Whitespace on the beginning or end of directory names is evil, but is permitted in some circumstances and we should allow for it. This also silently discards extraneous fields after the second field, matching the behavior of earlier versions of dirvish. I built a test script and fed it a few examples: ---------------------- input>AA BB< srctree>AA< aliastree>BB< ---------------------- input>AA BB CC DD< srctree>AA< aliastree>BB< ---------------------- input>AA\ BB CC DD< srctree>AA BB< aliastree>CC< ---------------------- input>AA\ BB CC\ DD< srctree>AA BB< aliastree>CC < ---------------------- input>AA\ \ BB\ \ CC < srctree>AA < aliastree> BB CC< ---------------------- input>< ERROR: no source tree defined ---------------------- Ponder this patch. There is probably an easier way. I plan to put it into the 1.3.X, which I have neglected for a long time but plan to work on this weekend. I could use some help, BTW, we need better programmers working on dirvish! Keith -- Keith Lofstrom keithl-cQ9U58XuirDQT0dZR+AlfA@xxxxxxxxxxxxxxxx Voice (503)-520-1993 KLIC --- Keith Lofstrom Integrated Circuits --- "Your Ideas in Silicon" Design Contracting in Bipolar and CMOS - Analog, Digital, and Scan ICs

Next Message by Date:

Re: No unexpired good images (what should I do?)

On Thu 16 Nov 2006, hanj wrote: > > I'm getting the "No unexpired good images" message during expire. Now, I've > seen reference to this in the past stating that this is due to error message > written in the summary file in that particular vault. Looking at the vault, > I'm seeing 'SUCCESS', so not sure as to why it can't expire it. This is my > 'base' image, and is out dated.. so not sure what to do. I'll provide a > snapshot of the directory and file sizes along with the summary file in that > particular vault. You're looking at that particular vault, and seeing "SUCCESS". That's exactly what dirvish is also doing. However, dirvish apparently can't find any *NEWER* images that are also successful, so it's leaving the last known good image intact. *That* is what the message is telling you. So, take a look at the newer summary files. PS: please wordwrap at around column 70, that makes it easier to quote properly... > comp zeus # du -sh * > 5.1G 20060920 > 1000M 20061012 > 338M 20061013 > 349M 20061014 That's a bit curious, as du -sh * will usually count each argument separately without taking into account hardlinks (that's only done within each argument). Now it looks like each next image was pretty incomplete, which of course would indicate that those images were in fact unsuccessful... To find out how much space is unique in each subdirectory, I usually do: du -h --max-depth=1 . Of course, you need GNU du, but you seem to have that as you also use -h. > Any ideas on what I need to do? I'm hoping that I don't have to start over > and re-init the vault. No, because you still have the one image that is declared successful. Paul Slootman

Previous Message by Thread:

Spaces in source files - redux

Almost two years ago, one of the very first patches submitted to the dirvish archive was a fix by Eric Mountain that allowed tree names to have spaces in them. Spaces in names are possible with unix, but are much more common with windows. On the wiki, R. Fried pointed out that Eric's fix breaks the regular behavior - it removes the last character of the source tree when there are both sourcetree and aliastree fields . R. Fried suggested removing the patch, which breaks the behavior Eric wants. Here is a patch that hopefully meets both needs, which is described at the end of http://wiki.dirvish.org/index.cgi?SpacesInSource The 1.2.0 behavior by jwschultz looked like so: ($srctree, $aliastree) = split(/\s+/, $$Options{tree}) or seppuku 228, "ERROR: no source tree defined"; $srctree =~ s(/+$)(); $aliastree =~ s(/+$)(); $aliastree ||= $srctree; The 1.2.1 patch by Eric (the current version) looks like: ($srctree, $aliastree) = split(/[^\\]\s+/, $$Options{tree}) or seppuku 228, "ERROR: no source tree defined"; $srctree =~ s(\\ )( )g; $srctree =~ s(/+$)(); $aliastree =~ s(/+$)(); $aliastree ||= $srctree; My suggested patched behavior looks like: $_ = $$Options{tree} ; s/(\s)/\\$1/g; # These two lines swap the escaped spaces with s/\\\\//g; # the nonescaped ones. s/(\\\s)+/$1/g; # replace multiple separators with only one ($srctree,$aliastree) = split /\\\s/ or seppuku 228, "ERROR: no source tree defined"; $aliastree ||= $srctree; For you non-Perl programmers ( which includes me 29 days per month ) the new behavior manipulates the spaces before the split, turning escaped spaces into simple spaces, and simple spaces into escaped spaces. Duplicate escaped spaces are removed, and *then* we do the split. Quite a cheesy hack, but it allows both the original "source and alias" behavior to work, and also the escaped whitespace behavior that Eric wanted to work. In addition, it allows alternate white space characters (like tab) to be preserved in weird directory names, and also allows escaped whitespace on the beginning or end of directory names. Whitespace on the beginning or end of directory names is evil, but is permitted in some circumstances and we should allow for it. This also silently discards extraneous fields after the second field, matching the behavior of earlier versions of dirvish. I built a test script and fed it a few examples: ---------------------- input>AA BB< srctree>AA< aliastree>BB< ---------------------- input>AA BB CC DD< srctree>AA< aliastree>BB< ---------------------- input>AA\ BB CC DD< srctree>AA BB< aliastree>CC< ---------------------- input>AA\ BB CC\ DD< srctree>AA BB< aliastree>CC < ---------------------- input>AA\ \ BB\ \ CC < srctree>AA < aliastree> BB CC< ---------------------- input>< ERROR: no source tree defined ---------------------- Ponder this patch. There is probably an easier way. I plan to put it into the 1.3.X, which I have neglected for a long time but plan to work on this weekend. I could use some help, BTW, we need better programmers working on dirvish! Keith -- Keith Lofstrom keithl-cQ9U58XuirDQT0dZR+AlfA@xxxxxxxxxxxxxxxx Voice (503)-520-1993 KLIC --- Keith Lofstrom Integrated Circuits --- "Your Ideas in Silicon" Design Contracting in Bipolar and CMOS - Analog, Digital, and Scan ICs

Next Message by Thread:

Re: No unexpired good images (what should I do?)

On Thu 16 Nov 2006, hanj wrote: > > I'm getting the "No unexpired good images" message during expire. Now, I've > seen reference to this in the past stating that this is due to error message > written in the summary file in that particular vault. Looking at the vault, > I'm seeing 'SUCCESS', so not sure as to why it can't expire it. This is my > 'base' image, and is out dated.. so not sure what to do. I'll provide a > snapshot of the directory and file sizes along with the summary file in that > particular vault. You're looking at that particular vault, and seeing "SUCCESS". That's exactly what dirvish is also doing. However, dirvish apparently can't find any *NEWER* images that are also successful, so it's leaving the last known good image intact. *That* is what the message is telling you. So, take a look at the newer summary files. PS: please wordwrap at around column 70, that makes it easier to quote properly... > comp zeus # du -sh * > 5.1G 20060920 > 1000M 20061012 > 338M 20061013 > 349M 20061014 That's a bit curious, as du -sh * will usually count each argument separately without taking into account hardlinks (that's only done within each argument). Now it looks like each next image was pretty incomplete, which of course would indicate that those images were in fact unsuccessful... To find out how much space is unique in each subdirectory, I usually do: du -h --max-depth=1 . Of course, you need GNU du, but you seem to have that as you also use -h. > Any ideas on what I need to do? I'm hoping that I don't have to start over > and re-init the vault. No, because you still have the one image that is declared successful. Paul Slootman