Summary: | emerge sync/webrsync destroys /usr/portage when tmp filesystem is full during sync | ||
---|---|---|---|
Product: | Portage Development | Reporter: | Andreas Kobara <abusch> |
Component: | Unclassified | Assignee: | Portage team <dev-portage> |
Status: | RESOLVED FIXED | ||
Severity: | critical | CC: | flash3001, mschulkind |
Priority: | High | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | All | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Bug Depends on: | 35535 | ||
Bug Blocks: |
Description
Andreas Kobara
2003-07-29 04:05:06 UTC
Portage team - any comments about this one? Err... that sounds more like a rsync bug then portage bug. Yes, might be a rsync issue... I did find this: http://groups.google.com/groups?selm=aqda8b%24oii%241%40FreeBSD.csie.NCTU.edu.tw&output=gplain It was suggested to change code in receiver.c to raise an ENOSPC instead of RERR_FILEIO at some places. This was Nov/2002, and now in 2004 with rsync 2.6.x we still have the same code/bug ? this is definitely not a release bug... Fact is that webrsync will kill your portage tree when running out of diskspace. IMHO rsync should handle this, at least we need a patch for gentoo to ensure that it doesn't get its portage-tree destroyed that way. (This also happens when the portage snapshot is incomplete as on 10242004, as rsync just does its job and deletes everything not inside the snapshot.) No way to handle when the snapshot is incomplete, and the md5sum of the snapshot is correct (in other words, if our mirrors are daft and release an incomplete snapshot, you're screwed and portage cannot detect it). Besides that, tar returns should be checked... for the sync refactoring, I know it is. For stable, the tar exit code is checked also. |