Attempting to install Text::CSV_XS via g-cpan yields the following error message: >>> Verifying ebuild manifests !!! Manifest file not found: '/usr/local/cpan-portage/perl-gcpan/Text-CSV-XS/Manifest' It looks like g-cpan is attempting to change the '_' to a '-', which in at least this case isn't correct. Reproducible: Always Steps to Reproduce: Just run `sudo g-cpan -i Text::CSV_XS'. Actual Results: g-cpan chokes and reports an error message. Expected Results: g-cpan should install dev-perl/Text-CSV_XS. cpan> m Text::CSV_XS Module id = Text::CSV_XS DESCRIPTION Fast 8bit clean version of Text::CSV CPAN_USERID HMBRAND (H.Merijn Brand <h.m.brand@xs4all.nl>) CPAN_VERSION 0.76 CPAN_FILE H/HM/HMBRAND/Text-CSV_XS-0.76.tgz UPLOAD_DATE 2010-10-13 DSLIP_STATUS RdcOp (released,developer,C,object-oriented,Standard-Perl) MANPAGE Text::CSV_XS - comma-separated values manipulation routines INST_FILE /usr/lib64/perl5/vendor_perl/5.12.2/x86_64-linux-thread-multi/Text/CSV_XS.pm INST_VERSION 0.72
The same error when generating the ebuild for the package Tree::DAG_Node. It generates Tree-DAG-Node instead of Tree-DAG_Node. It seems that this applies to all package names containing '_'.
I had a look at this. I managed to get the ebuild to work after hacking: /usr/portage/eclass/perl-module.eclass [[ -z "${SRC_URI}" && -z "${MODULE_A}" ]] && MODULE_A="${MY_P:-${P}}.tar.gz" to [[ -z "${SRC_URI}" && -z "${MODULE_A}" ]] && MODULE_A="${MY_P:-${P}}.tgz" and manually changing the name of the ebuild to have the _ in it. So, there seem to be two bugs: the .tgz and the _ they can probably both be fixed in /usr/portage/eclass/perl-module.eclass
(In reply to comment #2) > I had a look at this. I managed to get the ebuild to work after hacking: > > /usr/portage/eclass/perl-module.eclass > [[ -z "${SRC_URI}" && -z "${MODULE_A}" ]] && MODULE_A="${MY_P:-${P}}.tar.gz" > to > [[ -z "${SRC_URI}" && -z "${MODULE_A}" ]] && MODULE_A="${MY_P:-${P}}.tgz" This change would break a lot of other dev-perl ebuilds. The above is just the default. MODULE_A and a couple of other variables should be defined in the ebuild in this case it would be the next couple of lines however, see further below. EAPI=2 MY_PN="Text-CSV_XS" MY_P="${MY_PN}-${PV}" MODULE_A="${MY_P}.tgz" MODULE_AUTHOR="HMBRAND" inherit perl-module > and manually changing the name of the ebuild to have the _ in it. tove: there doesn't seem to be any consistency in which ebuilds have "_" -> "-" in the module name. Should g-cpan be changing that "_" -> "-"? Should we fix the following packages to get their _ back? crypt-des-ede3 => Crypt-DES_EDE3 OLE-StorageLite => OLE-Storage_Lite sdl-perl => SDL_Perl > So, there seem to be two bugs: > the .tgz and the _ > they can probably both be fixed in > /usr/portage/eclass/perl-module.eclass No, they are NOT bugs in the eclass. In the entire dev-perl, only 10 packages have .tgz extensions, 982 have .tar.gz. P.S. Why are you trying to use g-cpan for Text::CSV_XS when it's already in the main tree as dev-perl/Text-CSV_XS?
(In reply to comment #3) > P.S. > Why are you trying to use g-cpan for Text::CSV_XS when it's already in the main > tree as dev-perl/Text-CSV_XS? If g-cpan installs a module which depends on Text::CSV_XS, it will generate an ebuild for Text-CSV-XS. So it happens. > tove: there doesn't seem to be any consistency in which ebuilds have "_" -> "-" > in the module name. Should g-cpan be changing that "_" -> "-"? I don't know why g-cpan does this replacement. > Should we fix the following packages to get their _ back? > crypt-des-ede3 => Crypt-DES_EDE3 > OLE-StorageLite => OLE-Storage_Lite > sdl-perl => SDL_Perl I wanted to wait until we moved away from CVS before changing PNs. But for 3+2+1 packages we could do it now if it solves the problem.
(In reply to comment #4) > (In reply to comment #3) > > in the module name. Should g-cpan be changing that "_" -> "-"? > I don't know why g-cpan does this replacement. Fixed in Git now. > > Should we fix the following packages to get their _ back? > > crypt-des-ede3 => Crypt-DES_EDE3 > > OLE-StorageLite => OLE-Storage_Lite > > sdl-perl => SDL_Perl > I wanted to wait until we moved away from CVS before changing PNs. > But for 3+2+1 packages we could do it now if it solves the problem. Ok, potentially soon.
*** Bug 339889 has been marked as a duplicate of this bug. ***
I've covered the TGZ potion with bug 313049, where g-cpan now detects the extension (and supports a wider list).
0.16.3 released now. Renames postponed till later.