On Wed, Jun 03, 2020 at 08:55:39AM +0200, Mathieu Othacehe wrote:
> 
> Hey Jakub,
> 
> Thanks for fixing the cross-compilation, a full review this time :)
> 
> > +             (let* ((outdir (assoc-ref outputs "out"))
> > +                    (bindir (string-append outdir "/bin"))
> > +                    (mandir (string-append outdir "/share/man/man1"))
> > +                    (docdir (string-append outdir "/share/doc/hashcash-" ,version)))
> > +               ;; make install assumes /usr and doesn't provide a way to override it
> > +               (install-file "hashcash" bindir)
> > +               (install-file "hashcash.1" mandir)
> > +               (install-file "README" docdir)
> > +               (install-file "LICENSE" docdir)
> > +               (install-file "CHANGELOG" docdir)
> 
> I think you can set the variables INSTALL_PATH, MAN_INSTALL_PATH and
> DOC_INSTALL_PATH instead.

Unfortunately, the variables are defined with `=` instead of `?=`, so
the Makefile always overwrites them. We would also want to skip
installing the `sha1' tool like Debian does, since `sha1sum' is already
a thing.

> > +               #t))))))
> > +    (home-page "https://www.hashcash.org/")
> > +    (synopsis "Denial-of-service countermeasure")
> > +    (description "Hashcash is a proof-of-work algorithm, which has been used
> > +as a denial-of-service countermeasure technique in a number of systems.
> 
> You can remove "in a number of systems".
> 
> > +A hashcash stamp constitutes a proof-of-work which takes a parametrizable
> > +amount of work to compute for the sender.  The recipient can verify received
> > +hashcash stamps efficiently.
> > +
> > +This package contains a command-line tool for computing and verifying hashcash
> > +stamps.")
> > +    (license license:public-domain)))
> 
> I'm also concerned by this line in the Makefile:
> 
> --8<---------------cut here---------------start------------->8---
> # request static link of -lcrypto only
> LIBCRYPTO=/usr/lib/libcrypto.a
> --8<---------------cut here---------------end--------------->8---
> 
> We should maybe add "openssl" to the inputs and fix this variable
> (that's what Nix does).

From what I've read, openssl is not used by default, but only when a
target like `gnu-openssl' is used. I'll make a build with openssl and
benchmark which is faster...

Regards,
Jakub Kądziołka