Xavier wrote:

On Sun, Mar 8, 2009 at 7:16 AM, Allan McRae wrote:

...

Hi,

Am I missing something here or should these sorts of tests cause a failure:

devel_check() { ... # Also do a brief check to make sure we have the VCS tool available. ... [ $(type -p darcs) ] || return 0 ... [ $(type -p svn) ] || return 0

etc.

At the moment, these tests just seem useless to me. Making this an error would require a decent change in makepkg as at the moment devel_check() is called before dependency resolution is done. In fact, I don't think I have ever managed to build an SCM package in a clean chroot and this may be the cause.

If no-one points out something obvious I am missing, then I will file a bug report so this is not forgotten.

A SCM package should have correct makedepends anyway. Let's take a svn package, without svn makedepends. With a makepkg without devel_check, it would have failed in the build() function of the pkgbuild, where svn is called. With a makepkg with devel_check and without the above tests, it would fail in devel_check. So a mistake in the PKGBUILD (lack of makedepends=svn) would result in a failure in a makepkg internal function (devel_check) rather than in the pkgbuild build() function.

Yeah, I was just wondering if the devel_check should come after the installation of deps and makedeps so that if a VCS tool is missing, we can make it an error. At the moment, we can not update the pkgver during the first devel_check. So you end up with issues like this: ==> Making package: notify-sharp-svn 3009-2 i686 (Mon Mar 9 08:23:02 UTC 2009) ... ==> Installing missing dependencies... (subversion) ==> Entering fakeroot environment... ==> Determining latest svn revision... -> Version found: 3032 (note the different version there and how --forcever does not seem to have been used/work) ... ==> Finished making: notify-sharp-svn 3009-2 i686 (Mon Mar 9 08:25:36 UTC 2009) Hmmm..... while testing this I noticed this: ==> Validating source files with md5sums... ==> Validating source files with sha1sums... ==> Validating source files with sha256sums... ==> Validating source files with sha384sums... ==> Validating source files with sha512sums... Note that svn packages tend to have "md5sums=()" in the PKGBUILD. I wonder if that is still an issue? Allan

Xavier wrote:

On Mon, Mar 9, 2009 at 9:36 AM, Allan McRae wrote:

...

Yeah, I was just wondering if the devel_check should come after the installation of deps and makedeps so that if a VCS tool is missing, we can make it an error. At the moment, we can not update the pkgver during the first devel_check. So you end up with issues like this:

==> Making package: notify-sharp-svn 3009-2 i686 (Mon Mar 9 08:23:02 UTC 2009) ... ==> Installing missing dependencies... (subversion) ==> Entering fakeroot environment... ==> Determining latest svn revision... -> Version found: 3032 (note the different version there and how --forcever does not seem to have been used/work) ... ==> Finished making: notify-sharp-svn 3009-2 i686 (Mon Mar 9 08:25:36 UTC 2009)

I see, but in any cases, we won't be able to display this message where it is now :

...

==> Making package: notify-sharp-svn 3009-2 i686 (Mon Mar 9 08:23:02 UTC

unless we either remove the version, or move this message after installing missing dependencies (could look weird).

Here are the other things that happen between the current devel_check and the missing deps install, which relies on $pkgver : 1) check of an existing package (and install with -i) 2) creation of the source package

For 2, I don't see why it would be needed to bump the pkgver to the last one. For 1, it might be ok to not bump it either. This would make us aware that there is already a package built for the pkgver in the pkgbuild, so that we might want to just install that one. Otherwise, we can force with -f, which may or may not result in a different $pkgver after devel_update or devel_check, but well, that's what --holdver is for.

So it seems doable to delay devel_check, having the three points above in mind.

I am going to file a bug report so I do not forget about this. Given we may have a release soon, I don't think I want to make all the changes that this would require at the current time.

...

Hmmm..... while testing this I noticed this: ==> Validating source files with md5sums... ==> Validating source files with sha1sums... ==> Validating source files with sha256sums... ==> Validating source files with sha384sums... ==> Validating source files with sha512sums...

Note that svn packages tend to have "md5sums=()" in the PKGBUILD. I wonder if that is still an issue?

Ahah. What about [ ${#source[@]} -eq 0 ] && return 0 at the top of check_checksums ?

That would be a good solution. I can patch this up later. I have never actually looked at the check_checksums function before...