maemo.org Bugzilla – Full Text Bug Listing |
Summary: | osso-software-version-rx*4 dependencies prevent removing many unnecessary packages | ||
---|---|---|---|
Product: | [Maemo Official Applications] Settings and Maintenance | Reporter: | Josh Triplett <josh> |
Component: | Application manager | Assignee: | Quim Gil <quim.gil> |
Status: | RESOLVED FIXED | QA Contact: | application-manager-bugs |
Severity: | enhancement | ||
Priority: | Low | CC: | andre_klapper, barinowoleg2, eero.tamminen, josh, kemarken, maemo, maemo, maemoBugzilla-ugeuder, marcell, marius.vollmer, matt, quim.gil, rabelg5, texrat, urho.konttori |
Version: | 5.0-beta | ||
Target Milestone: | 5.0 (1.2009.41-10) | ||
Hardware: | All | ||
OS: | Linux | ||
URL: | http://www.internettablettalk.com/forums/showpost.php?p=229093&postcount=304 | ||
Bug Depends on: | |||
Bug Blocks: | 1798, 2435 |
> that extra 14MB or so (most of the time, the pre-installed-documentation-rx*4 > package accounts for the majority of an updates size) makes a big difference > in the installability of the update and the quality of the user experience. The file system compresses the files with LZO, so from the file system the PDF files take in reality <10MB[1]. Still a big problem though as JFFS2 keeps the whole file system structure in memory which means that it will use more memory when the file system is fuller (and significantly more for files that are open). System using more memory means that applications have less which means that they will abort() more easily[2]. [1] You can test this with something like: time tar -cv /home/user/MyDocs/.documents/ | lzop > test.tar.lzo du -k test.tar.lzo [2] Glib and everything based on it like Gtk will abort if memory allocation fails, there's not really much that can be done about that, all major UI toolkits work like that (handling alloc failures reasonably would increase code size significantly and explode amount of work needed. Not several times, but (several) magnitude(s) more if one wants to do it properly).
> The file system compresses the files with LZO, so from the file system > the PDF files take in reality <10MB. I think this pre-installed content might actually be the reason why the SSU fails for some people. As the file system is compressed, you cannot just calculate how much space an update requires, it needs to be tested and that requires quite a bit of iteration. I think we internally test this by filling the rootfs before SSU update is tried. However, that test results may not valid if user has manually removed some of things that are pre-installed and that SSU depends from. If they're not removed in the internal testing, those files would be just replaced on SSU update i.e. no additional space is needed. If user removed them, more 10MB space is used... As a conclusion: SSU shouldn't update any content that user can easily remove without removing the corresponding package. I think currently this concerns only pre-installed content.
> (In reply to comment #31) > > > Ville: Can this be considered for Fremantle? > > > > Certainly. We have an SSU-meeting on Friday, and the new usecases for > > Fremantle are on the very top of the agenda. Ville, any news here wrt Fremantle?