Interesting, although the proliferation of command line options begins again. Thoughts from anyone else?

I don't know how I feel about it. Shouldn't this, in essence, be a cleanup task (part of -c or perhaps -cc ?)?

Good point, it seems kind of odd to do this as a one-off operation when building a package. I'd be fine with closing without implementing.

I made this patch in order to safe space on disk for those who haven't big disk/haven't a lot a space left on disk, think about OOo/Kde/Firefox/Kernel/"another big netfile source".

Disc space is cheap...bandwidth is not for everyone. A pacman cache is likely a bigger disk user than one source package will ever be. Is running makepkg -c all that hard (or even deleting the source manually)?

Well my point was that makepkg -c doesn't cleanup the downloaded source and it doesn't sound like a good default really. Perhaps we could add on another -c for cleaning source as well? makepkg -cc?

Um...makepkg -C?

After talking this over with Dan, I will recant what I said.
I think this is already handled as well as it could be with a combo of setting SRCDEST and using makepkg -C

Ok, -C option cleans SRCDEST but in the meanwhile, netfiles are duplicated and during the duplication + extraction process of these files and you can running out of space on your disk.
Moreover, this patch is small and it's a KISS peace of code that doesn't deeply modify makepkg.

Why don't we fix the real problem then- the duplication? It would seem awfully counterintuitive to me to build a whole package, forget a configure flag, and then have to re-download the source because of it. Keep in mind that the disk space of the extracted source far exceeds that of the compressed source, that is the bigger problem...

I just don't see this being something that is needed by a large number of users- disk space is cheap, and deleting one tarball doesn't do a whole lot, even if it is 100MB (which means it will be way bigger when extracted).

Mm?? What you say is false, my patch doesn't delete netfiles just after decompression, it just avoids netfiles to be duplicated in SRCDEST (for those who want) and netfiles still exists in $startdir/src, that's all folks. So if makepkg process fails and you relaunch makepkg, netfiles aren't downloaded again..

Ahh, I see. Well why don't we address the real problem then? For any files downloaded, store them only in SRCDEST instead of SRCDEST and src/. This should be possible, no?

In fact, makepkg download those files in working directory src/ and then do a copy to SRCDEST, for local files makepkg copy SRCDEST to src/

IMO it should do in an opposite way:
download sources to SCRDEST, then extract to src/ (without copying)

Anyone want to make a patch against GIT makepkg incorporating Roman's good suggestion above?