wingrass: demolocation includes .svn/ cruft

[hamish]

the nightly build wingrass installers ship '.svn/' files for the demolocation.

perhaps somewhere needs to do 'svn export' not 'svn checkout'?

thanks, Hamish

[neteler]

Replying to hamish:

the nightly build wingrass installers ship '.svn/' files for the demolocation.

To better understand: is it a problem?

[hamish]

Replying to neteler:

Replying to hamish:

the nightly build wingrass installers ship '.svn/' files for the demolocation.

To better understand: is it a problem?

It does not cause breakage, but it should be removed.

Hamish

[hamish]

please don't ignore it. it's a 2 minute job and these things make us look unprofessional.

thanks, Hamish

[martinl]

Replying to hamish:

please don't ignore it. it's a 2 minute job and these things make us look unprofessional.

huh, "trivial" means ignoring? if it's a 2 minute job, feel free to fix it. Martin

[hamish]

.. it seemed to be going that way. Yes, I had a look at fixing it myself before filing the ticket, but couldn't see where in src:mswindows/ and addons:tools/wingrass-packager/ it was done. Any hints?

thanks, Hamish

[hellik]

Replying to hamish:

.. it seemed to be going that way. Yes, I had a look at fixing it myself before filing the ticket, but couldn't see where in src:mswindows/ and addons:tools/wingrass-packager/ it was done. Any hints?

​http://trac.osgeo.org/grass/browser/grass/branches/releasebranch_6_4/mswindows/GRASS-Installer.nsi.tmpl#L13

Helmut

[hamish]

so, IIUC, the packager does a 'svn up' on their machine into a local C:\osgeo4w dir, then the packaging script recursively copies from within that checked out source tree?

​http://trac.osgeo.org/grass/wiki/CompileOnWindows#CreatetheWinGRASSInstaller

Line 586 of GRASS-Installer.nsi.tmpl has:

File /r ${DEMOLOCATION_PATH}\*.*

how to restrict that, or afterwards recursively rm the 6 .svn/ dirs within it? or just forget anything fancy and remove them by hand after the copy:

RMDir /r "$GIS_DATABASE\demolocation/.svn"
RMDir /r "$GIS_DATABASE\demolocation/PERMANENT/.svn"
RMDir /r "$GIS_DATABASE\demolocation/PERMANENT/vector/.svn"
RMDir /r "$GIS_DATABASE\demolocation/PERMANENT/vector/mysites/.svn"
RMDir /r "$GIS_DATABASE\demolocation/PERMANENT/vector/point/.svn"
RMDir /r "$GIS_DATABASE\demolocation/PERMANENT/dbf/.svn"

​http://nsis.sourceforge.net/Delete_files_and_subdirectories

(I think that recursive remove is also needed for the extracted ms VC++ dll dir in Temp/, right now it gets left behind after install)

thanks, Hamish

[martinl]

Replying to hamish:

so, IIUC, the packager does a 'svn up' on their machine into a local C:\osgeo4w dir, then the packaging script recursively copies from within that checked out source tree?

see

source:grass-addons/tools/wingrass-packager/grass_compile.sh#L14​

Martin

[hellik]

Replying to hamish:

(I think that recursive remove is also needed for the extracted ms VC++ dll dir in Temp/, right now it gets left behind after install)

​http://trac.osgeo.org/grass/browser/grass/branches/releasebranch_6_4/mswindows/GRASS-Installer.nsi.tmpl#L1000

Delete "$TEMP\$ARCHIVE_NAME"
Delete "$TEMP\$ORIGINAL_UNTAR_FOLDER"

the downloaded files and folders should be deleted.

if e.g. vcredist_2005_x86.exe creates another folder with extracted files, shouldn't be this cleaning the os?

[hamish]

left over in ...\Local Settings\Temp\ are a install_msruntime dir from osgeo4w, with bin/ and etc/ within it, and an empty "Microsoft Visual C++ Runtime 2010 x86 Redistributible Setup_10.0.30319" dir, and matching 2010 .html and .txt log files.

wrt the 'install_msruntime' dir, does "Delete" work for folders?

Hamish

[glynn]

Replying to hamish:

the nightly build wingrass installers ship '.svn/' files for the demolocation.

I think that this isn't restricted to Windows; the "builddemolocation" target in the top-level Makefile simply copies the demolocation directory to dist.<arch> (using "tar c ... | tar x ...").

In 7.x, the demolocation directory has a Makefile which explicitly creates target directories and copies files with $(INSTALL_DATA).

Also: recent versions of the subversion client use a single .svn directory in the root of the working directory, rather than having a separate .svn directory in each subdirectory, so updating the subversion client would make this particular issue disappear.

[hamish]

attempt to remove the unpacked 'install_msruntime' dir in devbr6 r56001.

[hamish]

Replying to glynn:

Replying to hamish:

the nightly build wingrass installers ship '.svn/' files for the demolocation.

I think that this isn't restricted to Windows; the "builddemolocation" target in the top-level Makefile simply copies the demolocation directory to dist.<arch> (using "tar c ... | tar x ...").

that is correct, it's in all 'make install' for svn builds.

In 7.x, the demolocation directory has a Makefile which explicitly creates target directories and copies files with $(INSTALL_DATA).

since there's no Other.make in grass6 I'd just go with the cheap & dirty approach-of-least-change:

Index: Makefile
===================================================================
--- Makefile	(revision 55903)
+++ Makefile	(working copy)
@@ -389,6 +389,13 @@
 builddemolocation:
 	test -d ${ARCH_DISTDIR} || ${MAKE_DIR_CMD} ${ARCH_DISTDIR}
 	-tar cBf - demolocation | (cd ${ARCH_DISTDIR}/ ; tar xBfo - ) 2>/dev/null
+	-(cd ${ARCH_DISTDIR}/demolocation ; \
+		rm -rf ".svn" ; \
+		rm -rf "PERMANENT/.svn" ; \
+		rm -rf "PERMANENT/vector/.svn" ; \
+		rm -rf "PERMANENT/vector/mysites/.svn" ; \
+		rm -rf "PERMANENT/vector/point/.svn" ; \
+		rm -rf "PERMANENT/dbf/.svn" )
 	@ echo "GISDBASE: ${RUN_GISBASE}" > ${RUN_GISRC}
 	@ echo "LOCATION_NAME: demolocation" >> ${RUN_GISRC}
 	@ echo "MAPSET: PERMANENT" >> ${RUN_GISRC}

I notice that the last part creates a file called demolocation/.grassrc65, when it should perhaps be .grassrc6? That's named by Platform.make.in:

RUN_GISRC           = ${ARCH_DISTDIR}/demolocation/.grassrc${GRASS_VERSION_MAJOR}${GRASS_VERSION_MINOR}

(RUN_GISRC is used by e.g. Man.make at build time, with the GISRC= override set to the same variable, so it doesn't matter much what it's called)

?, Hamish

[hamish]

second call for comments on the cheap & dirty patch in comment:14.

[hamish]

patch committed in devbr6 and relbr64, but only affects 'make install', which the nsis installer doesn't use.

another attempt at cleaning it up wingrass committed in devbr6 r56107.

Hamish

[hamish]

wingrass7 was also copying over the .svn/ files (doesn't use 'make install', just copies from source dir), devbr6 fix for that merged to trunk in r56108.

I also notice in a wingrass7 install that a Makefile and grassrc.tmpl file have made it into $GISBASE/demolocation.

Hamish

[martinl]

Replying to martinl:

Replying to hamish:

please don't ignore it. it's a 2 minute job and these things make us look unprofessional.

huh, "trivial" means ignoring? if it's a 2 minute job, feel free to fix it. Martin

as you can see such "2 minutes jobs" are usually taking more then "2 minutes", currently stand-alone installer for G65 and G70 are broken

SetOutPath: "$GIS_DATABASE\demolocation"
File: Returning to: "."
File: "$DEMOLOCATION_PATH" -> no files found.
Usage: File [/nonfatal] [/a] ([/r] [/x filespec [...]] filespec [...] |
   /oname=outfile one_file_only)
Error in script ".\grass65\GRASS-Installer.nsi" on line 586 -- aborting creation
 process

[hamish]

Replying to martinl:

currently stand-alone installer for G65 and G70 are broken

SetOutPath: "$GIS_DATABASE\demolocation"
File: Returning to: "."
File: "$DEMOLOCATION_PATH" -> no files found.
Usage: File [/nonfatal] [/a] ([/r] [/x filespec [...]] filespec [...] |
   /oname=outfile one_file_only)
Error in script ".\grass65\GRASS-Installer.nsi" on line 586 -- aborting creation
 process

try r56130,1. Probably nothing to do with .svn/ additions, just me not understanding the nsis scripting language very well in a side-edit.

Hamish

[martinl]

Replying to hamish:

Replying to martinl: try r56130,1. Probably nothing to do with .svn/ additions, just me not understanding the nsis scripting language very well in a side-edit.

standalone installers are back (build 577) (1).

(1) ​http://wingrass.fsv.cvut.cz/

[hamish]

now working in 6.5 and backported to 6.4 in r56142.

other-cruft loose ends filed as #1966.

closing ticket.

Hamish