When building git from source, its configure script checks for the presence of
curl on the system being built and configures git-clone (and other git
commands) to avoid using curl if curl is unavailable. (see, e.g.,
git-clone.sh:359 from the source which reads:
if test -z "@@NO_CURL@@"
then
clone_dumb_http "$repo" "$D"
else
die "http transport not supported, rebuild Git with curl support"
fi
In the current Ubuntu package (1:1.5.4.3-1ubuntu1), that line reads
if test -z "",
presumably b/c the package was built on a system with curl installed. Run-time
detection seems out (transport.c also has curl-optional components) so perhaps,
like gitk, all the binaries/scripts that require curl support can be split out
into a separate package. git-clone-curl and git-clone-nocurl could be, for
example, selected between via alternatives.
When building git from source, its configure script checks for the presence of
curl on the system being built and configures git-clone (and other git
commands) to avoid using curl if curl is unavailable. (see, e.g.,
git-clone.sh:359 from the source which reads:
if test -z "@@NO_CURL@@" dumb_http "$repo" "$D"
then
clone_
else
die "http transport not supported, rebuild Git with curl support"
fi
In the current Ubuntu package (1:1.5. 4.3-1ubuntu1) , that line reads
if test -z "",
presumably b/c the package was built on a system with curl installed. Run-time
detection seems out (transport.c also has curl-optional components) so perhaps,
like gitk, all the binaries/scripts that require curl support can be split out
into a separate package. git-clone-curl and git-clone-nocurl could be, for
example, selected between via alternatives.