"some flexbuff has 2 out of 6 mount points totally full.
recording directory can still be created but no chunks can be
written. because of this vbs_rm does not recognize those
directories as being part of the recording and thus does not remove
them."
Fixed by always adding directories to a recording name and keep track
of wether it contained vbs chunks. Then later on if any of the dirs
contain vbs chunks then it was a vbs recording and all dirs are
removed.
Also if all directories are empty they will be removed.
git-svn-id: svn+ssh://code.jive.eu/code/svn/vbs_fs@58 5b2df0cb-e17b-44c8-90c8-480d528b6e0d
is deprecated -> warning -> error -> no compile
Rewritten to use readdir(3).
* found issue with merging multithread search results:
duplicate/failing inserts were not caught by
std::copy(this_thread_results.begin(), this_thread_results.end(),
insert_iterator(totals))
Rewritten to manual for(...) loop copying like an
animal. But at least we can check each duplicate insert
git-svn-id: svn+ssh://code.jive.eu/code/svn/vbs_fs@57 5b2df0cb-e17b-44c8-90c8-480d528b6e0d
variable that is not even there in that context. So obviously,
this code was never really compiled, I guess ... Woopsie!
git-svn-id: svn+ssh://code.jive.eu/code/svn/vbs_fs@55 5b2df0cb-e17b-44c8-90c8-480d528b6e0d
miserable on Mk6 - no data can be read /at all/.
Investiation reveals ::mmap(2) failing with EINVAL. Out of the four possible causes for
EINVAL 'offset must be a multiple of the system pagesize as returned by getpagesize(2)'
is the most likely. Indeed, reworking to ::mmap() with an offset which is a multiple
of the system pagesize now works. Thus also the mapped size need to be adjusted and
when munmap(2) is called the original, page-aligned, pointer needs to be reconstructed.
git-svn-id: svn+ssh://code.jive.eu/code/svn/vbs_fs@53 5b2df0cb-e17b-44c8-90c8-480d528b6e0d
getting a chunk's data. Random access now does not suffer from
having to read through the whole file.
git-svn-id: svn+ssh://code.jive.eu/code/svn/vbs_fs@52 5b2df0cb-e17b-44c8-90c8-480d528b6e0d
be passed pattern(s) ("-I ..") to index only recordings matching those
pattern(s). Thanks to Jun Yang for originally suggesting this.
git-svn-id: svn+ssh://code.jive.eu/code/svn/vbs_fs@51 5b2df0cb-e17b-44c8-90c8-480d528b6e0d
and "." in their names as equals (don't show up at all). Now they do.
git-svn-id: svn+ssh://code.jive.eu/code/svn/vbs_fs@49 5b2df0cb-e17b-44c8-90c8-480d528b6e0d
did not handle reading past end-of-file and then seeking back to
beginning correctly. Now it does
git-svn-id: svn+ssh://code.jive.eu/code/svn/vbs_fs@48 5b2df0cb-e17b-44c8-90c8-480d528b6e0d
formatted like <experiment>_<station>_<scan>
* Changed to documentation to reflect transparent Mark6/FlexBuff support
git-svn-id: svn+ssh://code.jive.eu/code/svn/vbs_fs@46 5b2df0cb-e17b-44c8-90c8-480d528b6e0d
permission then scanning for vbs recordings would throw an exception
git-svn-id: svn+ssh://code.jive.eu/code/svn/vbs_fs@45 5b2df0cb-e17b-44c8-90c8-480d528b6e0d
(a future version may re-acquire them, though)
* changed text to reflect that it will build/install
three utilities: vbs_ls, vbs_rm and vbs_fs
git-svn-id: svn+ssh://code.jive.eu/code/svn/vbs_fs@43 5b2df0cb-e17b-44c8-90c8-480d528b6e0d
- more efficient; less memory
- supports arbitrary mount point patterns ("-R ..." option);
"-6" and "-v" (default, FlexBuff mountpoints) are just shorthands
for adding predefined patterns to the path list
- supports MIT Haystack dplane v.2 format / EVN FlexBuff
transparently - all recognized recordings are presented
as single files
- retains uid, gid and permissions of individual recordings
- support 'quick' startup - only catalog recording names.
Recording is fully indexed when opened for the first time.
Note: application must open recording twice/use e.g. touch(1)
to trigger indexing the recording of interest.
[The touch(1) will yield an error but it will trigger
indexing fair enough]
git-svn-id: svn+ssh://code.jive.eu/code/svn/vbs_fs@41 5b2df0cb-e17b-44c8-90c8-480d528b6e0d
vbs_ls and vbs_rm have been changed to the effect that that
isn't used anymore
git-svn-id: svn+ssh://code.jive.eu/code/svn/vbs_fs@39 5b2df0cb-e17b-44c8-90c8-480d528b6e0d
* added '-F' flag [cf. man ls(1)]
* output of missing chunks updated a bit
* honour order of 'sort by time' and 'sort by size' on cmdline
git-svn-id: svn+ssh://code.jive.eu/code/svn/vbs_fs@30 5b2df0cb-e17b-44c8-90c8-480d528b6e0d
* human readable size print algorithm simplified
* if listing recording names only, do not analyze all chunks
git-svn-id: svn+ssh://code.jive.eu/code/svn/vbs_fs@25 5b2df0cb-e17b-44c8-90c8-480d528b6e0d