(old)Puppy Linux Discussion Forum

amigo wrote:The backup location is set with the --root option to sentry(use --backup=yes to turn the feature on), or you can do it like this:
export INSTW_ROOTPATH=/path/to/backups
# turn the backup feature on
export INSTW_BACKUP=1
just give those two command in the terminal where you are running from, before running 'sentry make install'.

Awesome.

I have put zero time into a GUI for src2pkg. The command-line interface is so simple, and yet powerful -a GUI would only slow most users down. I did implemet the drag-n-drop functionality, though, and use it every day. I am ROX user, so for a long time dnd would only work with ROX-filer, but it now works with KDE/XFCE/GNOME as well.
Still, I sympathize with the desire for a GUI and will help you work it out -if you don't drive me mad in the process LOL. Most srcpkg users just use it in the simplest of fashions and it will work without problems for at least 80% of anything you throw at it. But for more complex builds, it can use build scripts which guarantee the repeatability of your builds and the scripts are arch-independent. You can use the same script to compile on other arches -even multi-lib distros. So when you have the hang of using it, you can easily copy over your build scripts to that puppy-ppc or puppy-x86_64 machine and rebuild every package with the same ease. there are example scripts included and even group-build scripts which show how to rebuild a whole tree full of many sources with a single command.

How lovely. As far as a GUI, it would mostly be a GUI-Wrapper. PCompile itself is mostly just an argument interpreter and wrapper for src2all, which has been integrated in now, so it is actually a script, but it doesn't do all that much. Anyhow, it's just that the GUI gives a chance for the less-command-line-oriented puppy users to say "Hey I want this program. Bam. Done."

That's what we were aiming for with pcompile.

This does sound nice though.
Got your email, will be testing and playing with it soon.

Techno, in case you didn't get it earlier, let me know when you're done with dump2pet and dir2pet updates and cleanups so they behave nice.

I'll finish up with pcompile to 0.2.0 for making the interface and all that work nice. Hopefully the comments should make things work well.

After that I'm looking at src2pkg -- if I can EVER get it to install properly, lol. It may be a good choice for a unified option. Trying it out will give an idea.

gedrean, I've sent you a new copy with the changes which should make src2pkg install and setup properly.

Hi amigo,

I installed src2pkg yesterday in "dpup" (yes, yet another branch) and I discovered I needed 'makepkg'... which I found in none other than big_bass 'slaxer_pup', (incidentally, the ver of src2pkg in slaxer_pup was 1.9.8, now have 1.9.9) and also 'upgradepkg' (is that a 10 yr old script? Testament to Pat's coding), it also complains about something else but basically works. It seems to run through the '--setup' routine wvery time I try it though.

Maybe you need those files Gedrean?

Cheers

01micko, this has been fixed so that running 'src2pkg --setup' will create and install a pet package.

email me and I'll send you a copy of the latest version -I am only distribtuing it privately right now while still developing the new *.pet and *deb support.

Gedrean - I initially just modified dir2pet to use basic xdialogs in place of the command line sections, but felt that it was still too annoying. I am currently working on combining the dir2pet and petspecs programs into one GUI with mouseover descriptions for each field. I just haven't had a lot of time to work on it lately due to 12+ hr days recently and 3 hrs of commute time. I am close to 50% complete, but with the amount of rewrite that I am doing it will need some testing.

technosaurus wrote:Gedrean - I initially just modified dir2pet to use basic xdialogs in place of the command line sections, but felt that it was still too annoying. I am currently working on combining the dir2pet and petspecs programs into one GUI with mouseover descriptions for each field. I just haven't had a lot of time to work on it lately due to 12+ hr days recently and 3 hrs of commute time. I am close to 50% complete, but with the amount of rewrite that I am doing it will need some testing.

That is WICKED.

Keep in mind, too much combination could be very difficult to work with, as users that just want to create a petspecs file don't want to build to a pet just yet.
Remember also, gtkdialog has a "tooltip" option (undocumented) for most controls, so you don't need to hardcode most mouseover comments. And to be honest, you only really need it for labels and certain buttons.

Good luck and I look forward to playing with your new Brundlefly.

I am planning to put 2 checkboxes for keeping existing pet.specs file and .desktop files - if one is found it will default to true. This will prevent accidentally overwriting a desktop file with locale support with a basic one.

some progress on gui - still needs quite some cleanup, but I need some sleep right now

TODO: tooltips, find executables, fileselection for icons,split out the packaging part to enable adoptpet and adoptpxt (maybe tgz, txz others later?)

Oh, techno, once you're ready the pcomp_script can be taken out of debug by removing the extra gtkdialog3 calls (they both call the invalid package message, around the make and make install and all those commands) - at this point I'm not really pushing a huge amount of my time into looking at this thread every day because I'm kinda at a stopping point for pcompile - unless you start finding new ideas. Adopt is a neat name though

Good luck with all this. I'll keep this thread on my watchlist but I'm gonna focus on other things, so if you have a request or idea go ahead and post it here - I should be able to play with it.

As it is, I'm right now trying to figure out how src2pkg is creating its slack.desc files -- because it's pulling that data from somewhere and I don't know where... Maybe amigo will see this and answer the unasked but implied question therein.

Those details could DEFINITELY be useful for pre-filling in other files like .desktop or such.

Aha! You've discovered one of the nicer features of src2pkg. It seems such a simple thing to create a slack-desc file. But it consumes a lot of time. Before creating the routines which 'mine' the description, writing the desc file usually tooik more time than the whole rest of the process of creating a package. so, even though it was a bit daunting at first, I started looking for ways to pull a description from existing files.

What you are seeing is probably pulled from an rpm *.spec file included in the sources, or from a debian/control file in the sources. These are the two most common and best places to get descriptive info. And for that very reason I usually use the debian patch for a given source tarball, or use a *.src.rpm I can find one. Often, after a first build which gets me the description, I thn stop using the patch or use the tarball instead of the src.rpm -I hate writing desc files.

src2pkg can use text found in *.spec, debian/control, *.lsm, PKGINFO or PKG-INFO(from python modules sources), description-pak files like used by checkinstall(some distros like vector linux use these) or from a NAME.txt file which accompanies a slackware binary package.

Somer real tricky code in there for wrapping the text and each type of description file needs its' own parser -but it's all done in pure bash so it is faster than calling externals.

You may have also noticed that src2pkg will discover what kind of license is being used by the software and write that info to the desc file, as welll as the URL of the homepage

The code which wraps the text allows you to change the line-length if you want and you can also set the number of lines to use in a description file -although the standard length for slack packages is 9-13

As some users have suggested, it would be nice to be able to simply 'read' the README and get the info, but 'simply' is simply not that simple! Anyway, The whole routine is one of the most complex and *lovely* routines in src2pkg. I don't know whether to be more proud or pleased as it saves me loads of work.

I figured that was the case. I was MORE impressed by the fact that it found out the license for SDL was LGPL -- but also that it was LGPL 2.1 -- and I couldn't find the 2.1 designator ANYWHERE in source files or text files anywhere.

So I don't know HOW it knew the LGPL version number. Did it just guess, or did it figure out based upon some of the wording in the license?

Yep, it greps for key phrases typical of each license type. Handy. huh?

A couple of quick pointers since you are up and running:

Use the option -Z? to reduce package size:
-Z1 uses bzip2 on man-pages
-Z2 uses upx-ucl, upx or exepak to compres binaries
-Z3 will link the License of the package to a common location (each package contains and installs the license, but subsequent installation which use the same license overwrite the same file.
-Z4 creates an archive of all docs.
Edit: The -Z options are cumulative: -Z4 does all of -Z1, -Z2 and -Z3.

Open /etc/src2pkg/src2pkg.conf and scroll down to line ~230
Uncomment and fill in all those extra compiler flags you want to use:
[[ $EXTRA_FLAGS ]] || EXTRA_FLAGS="-pipe"

If you want to use a bunch of extra options to configure scripts, uncomment and fill in line ~289:
AUTO_CONFIG_OPTIONS="sysconfdir localstatedir"
then use the -ACN option when running src2pkg. Note that src2pkg will examine the configure to script to see if these options are actually available and only those which are will be added to the config options.

Another note: you don't ever need to specify libdir, mandir or infodir as these are handled automatically. well, you can specify libdir, but you should never need to. Since src2pkg handles man-pages and info files -correcting their location when wrongly installed, it doesn't matter where they get installed -src2pkg will fix them according to which FSH_POLICY is in force.

You may find it nicer to uncomment these lines:
[[ $SOURCES_DIR ]] || SOURCES_DIR="$CWD"
[[ $SRC_BUILDS_DIR ]] || SRC_BUILDS_DIR="$CWD"
[[ $PKG_BUILDS_DIR ]] || PKG_BUILDS_DIR="$CWD"
[[ $PKG_DEST_DIR ]] || PKG_DEST_DIR="$CWD"
[[ $BACKUP_DIR ]] || BACKUP_DIR="$CWD"
They cause everything to be built in the current directory which makes it *much* easier to examine the package tree, or go into the sources for whatever reason.

If you have sound, you may wish to uncomment line ~347:
AUDIO_NOTIFICATION=SAY
Set to BEEP or PLAY or, if you have flite or festival (not likely) you can use SAY. Handy when you are watching something else while building packages in the background.
You can get a build script and flite sources here:
http://distro.ibiblio.org/pub/linux/dis ... flite-1.3/

Here is a full pet of the cumulative results. more, smaller scripts for building pets and pxts (requires xz and tar >= 1.22)

Fixes:
Fixed expansion for bad tarballs (ex. SDL)
Replaced ../local/pcompile/...... with /usr/local/pcompile/.... so you don't need to run it from within /usr/somedir
New GUI for making pet packages for the new format pet.specs
tweaks to default options

New utilities
adopt - fixes a dir for building a pet (used by adoptpet and adopt pxt)
adoptpet - fixes dir and actually builds pet
adoptpxt - fixes dir and actually builds pxt
dir2pet2 - assumes dir has been "adopted" and turns it into a pet
dir2pxt - assumes dir has been "adopted" and turns it into a pxt (uses xz compression instead of gz like pet packages do)
dumpdir - skips the "adopt" gui and assumes all is good
dmp2pet - runs dumpdir and dir2pet (for DEV, DOC, NLS or known good packages that already have proper menu entry and pet.specs)
dmp2pxt - runs dumpdir and dir2pxt (same as dmp2pet but better compression)

typo in gzip tarballs - here is the fix:

technosaurus wrote: Replaced ../local/pcompile/...... with /usr/local/pcompile/.... so you don't need to run it from within /usr/somedir

I //KNEW// I forgot a debug setting.

New GUI for making pet packages for the new format pet.specs
tweaks to default options

Wicked. I can't wait to play!

Alright, from reviewing the code, so far I have the following comments...

In the EXTENSIONS case statement, at one point we check if it's "tar" (the case), then we check if it's "tar" (the if) and if so, we make it "tar"... the if statement is really only necessary to check if there's another extension BEFORE the final extension... such as in the case of .tar.gz, or .tar.bz2 ... so that if statement in there is superfluous and can be removed, saving us a grep.

I noted the removal of the Invalid_Package gtkdialog, replacing with an xdialog -- this is nice as it uses less RAM, is faster, and has less space in the script. Nice going.

Also, it would be ABSOLUTELY lovely to get the source directory from TAR, in case it doesn't match, but the problem with that is that ZIP can contain multiple directories, and if we're including ZIP support, to be honest we have to hope/assume that the maintainers of the source code followed standards and named the tarball or zip arch the same as their source directory -- if not, to be honest, we may end up having this run on non-source tarballs, creating unexpected results when the user, say, selects the wrong tarball. We don't really want to screw that up, and the output of TAR isn't gonna tell us much, unless we want to build a rather complex parser for TAR output, which has to determine if the tar is even a directory, as it could just be a bunch of files! While it's not our job to clean up after the user, it's not really necessary to assume the code maintainers did things wrong... that's just my opinion, and really shouldn't add up to two grains of salt, but still. If you wanna do that, be my guest.

The extra invalid dialog calls (I see the again why? ) were to use as debug, so that if errors occurred from make, they weren't erased by dir2pet's horrible clear command.

The use of the XDialogs all over to replace my troubled GTKDIALOGS was brilliant. Thanks.

All in all, very wonderful.

Though I //am// wishing that at the end, the specs file remained in the package dir, so we could see the specs file without extracting out an extra copy of the pet package. <G>

EDIT: Oooh! Oooh! Move the specs and pinstall.sh to the rtdir!

Oops, run through gave me a problem...

It gave me a message "/.pet created".

And no SDL pet. <G> Trying again but I think there may be errors in adoptapet...

EDIT: I have a 0 Byte pxt file! That help?

Okay, changed dir2pxt in the final section (where it builds pets) to dir2pet2 -- it built, but gave the same message (/.pet created) -- however it VERY QUICKLY made that box go away (I didn't click anything) and then exited with the whole "we're done" message.

However it DID create a pet - so I think there may be some tweaking with dir2pxt to be done.

Particularly as I don't think I have xz installed -- maaay want to have a check if xz exists, if so do the pxt -- or may want to have a checkbox in pcompile that is auto-unchecked that says "make PXT - smaller but requires xz and updated pxtget on end-user's machine" or something like that... tooltips could be useful!

gzipped tarballs - I may download the latest gzip and see if it supports leaving the existing file in place since some tar.gz files (specifically the ones for SDL and friends) are not recognized by tar for whatever reason (this is why I decided to add plain tar support in the first place)

I also changed some of the default values to false so that it doesn't stop a configure script due to a missing argument

Forgot to mention pxt support (xz compressed pets) currently set to use default compression through tar, but may get better results by using xz separately using -ze9

'adopt' needs a better method of doing the menu category AND I kinda screwed up by setting the default value to the program name (which will never be a useful category) - unfortunately there is no "Miscellaneous" category

Somewhat unrelated but concerning tar:
I also notice that the old tar supported the -y option for bzip, but using tar 1.22 (which supports xz) this is no longer available - now only -j works for bzip/bz2 and -J is for xz. I think I fixed all -y occurances (actually -xyf) but hope that this tar doesn't break any compatibilities for other scripts. I am going to take a look at the Xarchive wrapper scripts since it hasn't been updated for a couple years now.