David Sainty <david.sainty%gmail.com@localhost> writes: > On 14/09/15 03:03, Joerg Sonnenberger wrote: >> Module Name: pkgsrc >> Committed By: joerg >> Date: Sun Sep 13 15:03:57 UTC 2015 >> >> Modified Files: >> pkgsrc/mk/fetch: bsd.fetch-vars.mk fetch.mk >> >> Log Message: >> Introduce IGNORE_INTERACTIVE_FETCH to allow trying to fetch files from >> the override site, even for INTERACTIVE_STATE=stage or FETCH_MESSAGE. >> Ignore all backup sites as they won't have the distfile anyway, the >> fetch wouldn't be interactive in first place otherwise. > > This breaks for PKG_RESUME_TRANSFERS=yes. > > This (head scratcher!) problem appears to be because > PKG_RESUME_TRANSFERS=yes forces do-fetch-file to run unconditionally, > even if the distfile already exists. > > That normally works, because fetch checks the checksums and exits > successfully; or otherwise restarts the download. > > "Exits successfully" is what should have happened below (because the > distfiles are already downloaded), but "fetch" never gets that far, > because it demands at least one source site, which just happened to be > the MASTER_SITE_BACKUP that is now inhibited for interactive fetch > packages. So instead we get a fatal usage error for the empty list of > download sites. > > Probably fetch should allow zero fetch locations, given the current usage. > > Bonus bug: this is already past the interactive fetch message, so > actually someone with a corrupted/partial/zero length distfile will > never see the interactive fetch instructions either. > > > [~/pkgsrc/lang/oracle-jre8] % make install > => Bootstrap dependency digest>=20010302: found digest-20121220 > ===> Checking for vulnerabilities in oracle-jre8-8.0.60 > => Fetching jce_policy-8.zip > usage: fetch [-c] [-d dir] [-f distinfo] [-p hook] [-r] [-v] file site ... > *** Error code 1 > > Stop. > bmake: stopped in /home/pkgsrc1/pkgsrc/lang/oracle-jre8 > > > .... and work around the problem ... > > % make PKG_RESUME_TRANSFERS=no install > => Bootstrap dependency digest>=20010302: found digest-20121220 > ===> Checking for vulnerabilities in oracle-jre8-8.0.60 > => Checksum SHA1 OK for jce_policy-8.zip > => Checksum RMD160 OK for jce_policy-8.zip > => Checksum SHA1 OK for jre-8u60-linux-i586.tar.gz > => Checksum RMD160 OK for jre-8u60-linux-i586.tar.gz > ===> Installing dependencies for oracle-jre8-8.0.60 > ... Install etc etc etc... Does one of you have a proposed fix? Or should we back out the above change, for redoing post freeze? It sounds like it might not be too hard/scary to fix - would be good to get a patch to review to tech-pkg.
Attachment:
pgpyrqnm1l5wL.pgp
Description: PGP signature