On Sat, 31 Aug 2013, Jed Brown wrote: > Satish Balay <[email protected]> writes: > >> Candidate for maint: refs/remotes/origin/jed/avoid-getpwuid Fri Aug 30 > >> 18:10:37 2013 -0700 > >> 6c8deb0 Sys: fix PetscGetFullPath expansion of ~ and ~user, ensure NULL > >> termination > >> 45082d6 Sys: avoid getpwuid in file path routines > >> 311ae19 Sys: add failing test expanding ~/file-name and ~user/file-name > >> > > > > BTW: perhaps this is a master only change? [because now ~/.petscrc won't > > work for some folks?] > > I started it from 'maint' because I wanted to have the option. Do we > know of systems for which getpwuid returns good information about the > home directory, but for which HOME is not set? Maybe we'll find out if > people are trying 'next' on funny machines. It would be nice for 3.4.3 > to not have this possible problem on BG.
I suspect it will be tested a bit in master - but wierd machine usage might primarily be from maint. [i.e release] > Note that PetscGetFullPath was obviously broken before: > > commit 311ae19d0f5ae5259a6236e6f4ce4509eaf01c8f > Author: Jed Brown <[email protected]> > Date: Fri Aug 30 15:24:26 2013 -0700 > > Sys: add failing test expanding ~/file-name and ~user/file-name > > > $(realpath ~/file-name) != > /home/jed/petsc/src/sys/examples/tests/~/file-name > > $(realpath ~jed/file-name) != > /home/jed/petsc/src/sys/examples/tests/~jed/file-name > > > Satish, do you think we could create a dashboard similar to Karl's > nightly summary that lists all these merge candidates? Would that be > too dangerous because it's hard to allow comments on whether a candidate > is viable? I can addin a weekly cronjob that sends an e-mail to petsc-dev - this would allow comments.. [we would have to update the script to remove the branches that started on or before v3.4 from the listing - or delete these branches - provided they are done/merged to master.] Satish
