shell scripts: %TMP% namespace collision

[hamish]

Hi,

• MS-Windows sets %TMP%
• MSys translates existing DOS variables into shell variables
• %TMP% becomes $TMP
• many shell scripts use TMP=`g.tempfile -p $$`
• many shell scripts try to remove $TMP* files even if they don't exist.
• potential ka-boom if g.tempfile never ran

I just triggered this with a typo: m.proj hlep

should we unset TMP at the start of any script which uses it?

d.polar
db.dropcol
i.fusion.brovey
m.proj
r.in.wms
r.mask
r.plane
r.regression.line
r.univar.sh
v.colors
v.db.dropcol
v.db.renamecol
v.db.univar
v.in.garmin
v.in.gpsbabel
v.in.lines
v.in.mapgen
v.in.wfs
v.out.gpsbabel
v.rast.stats
v.univar.sh

or should we do that as part of WinGrass's Init.sh?

Hamish

[glynn]

Replying to hamish:

should we unset TMP at the start of any script which uses it?

Use a different variable.

or should we do that as part of WinGrass's Init.sh?

No. That would cause problems for anything which expects it to be set. As would changing its value.

[hamish]

d.polar, db.dropcol, i.fusion.brovey, and m.proj done in 6.5svn with r40686.

r.in.wms, r.mask, r.plane, r.regression.line, r.univar.sh done in 6.5svn with r40687.

v.* still needs it.

[hamish]

v.* done in 6.5svn with r40689 and r40690.

[hamish]

backport this to relbr6.4 before or after rc6?

Hamish

[neteler]

Replying to hamish:

backport this to relbr6.4 before or after rc6?

Please (let's) do it before RC6 to avoid a RC7...

Markus

[glynn]

Replying to hamish:

v.dissolve and v.in.mapgen weren't listed but also use $TMP. I've fixed these in 6.5 with r41338.

[neteler]

Bulk-backport of r40686, r40687, r40689, r40690, r41338 in GRASS 6.4 as r41340.

Markus