Re: rpmlib installation qualms

["Jeff Johnson" <n3npq jbj gmail com>]

On 11/22/06, R. Tyler Ballance <tyler bleepsoft com> wrote:
> Just out of curiosity, is the
> int rpmInstall(rpmts ts, struct rpmInstallArguments_s *ia, const char
> **fileArgv);
>
> function just there to tease me? :)
>
> I'm calling it with a valid transaction, NULL rpmInstallArguments, and then
> a valid string (current working directory, or is an actual full path
> necessary?) and it SIGSEGVs every time.

There are 2 problems there. You need to pass non-NULL rpmInstallArguments,
and the 3rd argument is an array of string ptrs, not a string.

   struct rpmInstallArguments_s *ia = alloca(sizeof(*ia));

   memset(ia, 0, sizeof(*ia));

    ia->installInterfaceFlags = INSTALL_INSTALL|INSTALL_HASH;

is equivalent to -ih from the CLI, and is likely the minimal necessary
for calling rpmInstall.

> Are the install arguments necessary? where else might the SIGSEGV be
> stemming from? what is the flying weight of an unladen sparrow?

Necessary, yes.

Unladen sparrows dissipate waste energy through segmentaion fault droppings.

> I've seen references to old mailing lists questions that seem to be from
> before rpmInstall() was created, and have relatively complicated sample code
> for installation in the thread, but rpmInstall (and rpmInstallSource()) seem
> to be what is needed, if they'd just work :)
>
> Also, how are conflicts handled, or dependencies, etc? I want to make sure
> my code accommodating of some of the error cases rpmlib might present.

Almost all bits to control transaction operations passed to rpmInstall() are in
   ia->interfaceFlags
   ia->transFlags

73 de Jeff