* Peter Rosin wrote on Wed, Jun 23, 2010 at 06:12:43PM CEST: > While I didn't see any regressions, I did notice that it didn't > go too well with MSVC either. So, I dug around and found a > forgotten gem [1], and updated it to apply on top of the previous > patch. Result attached. > > So, I'm asking if this too can be commited, if the testsuite > finds no regressions... > > [1] http://lists.gnu.org/archive/html/libtool-patches/2008-09/msg00000.html
Okay if you replace test -e with test -f in this and prior patch, and if the testsuite addition exposes the issue, and the test passes on one unixy system. I'm not happy about the exeext normalization here, that probably belongs alongside the rest of the exeext hanling in ltmain. OTOH, there, no case-insensitive handling is done. Why should it be needed here? Do you expect package authors to write .EXE in their makefiles? Note that you aren't picking up file names from the system here; right? Thanks, Ralf
