Panu Matilainen [Mon, 28 Sep 2009 08:09:32 +0000 (11:09 +0300)]
OSGi dependency generator fixes from Alphonse Van Assche (ticket #101)
- This patch fix some minor bug, remove unused code and increase drastically
the perf by threading execution of tasks and using zip command instead of
jar to get meta-data
Panu Matilainen [Thu, 24 Sep 2009 10:57:55 +0000 (13:57 +0300)]
Enable subtyping on the rest of our type-objects
- not very useful atm as various places return hard-wired built-in types
Panu Matilainen [Thu, 24 Sep 2009 10:40:44 +0000 (13:40 +0300)]
Eliminate all custom tp_free() type methods
- tp_free()'s purpose is only to free up the memory used by the python
object structure, cleaning up our own allocations belongs to tp_dealloc()
Panu Matilainen [Thu, 24 Sep 2009 10:19:20 +0000 (13:19 +0300)]
Enable subtyping on rpm.ts class
Panu Matilainen [Thu, 24 Sep 2009 10:05:36 +0000 (13:05 +0300)]
Call generic python setattr if not scriptFd
- yet more preliminaries for subtyping
Panu Matilainen [Thu, 24 Sep 2009 10:04:34 +0000 (13:04 +0300)]
tp_setattro() is supposed to return -1 on errors
Panu Matilainen [Thu, 24 Sep 2009 09:52:32 +0000 (12:52 +0300)]
Call (sub)type tp_free from destructors
- more preliminaries for subtyping
- remove pointless NULL checks
Panu Matilainen [Thu, 24 Sep 2009 08:42:17 +0000 (11:42 +0300)]
Make object allocation type agnostic
- pass (sub)type down to wrappers
- call subtype tp_alloc() instead of PyObject_New()
- preliminaries for allowing subtyping
Panu Matilainen [Wed, 23 Sep 2009 14:44:07 +0000 (17:44 +0300)]
Remove unnecessary header type checking
- the type is already validated by PyArg_ParseTupleAndKeywords(),
no need to doublecheck
Panu Matilainen [Wed, 23 Sep 2009 11:41:30 +0000 (14:41 +0300)]
Remove the broken tp_print method from spec
Panu Matilainen [Wed, 23 Sep 2009 11:21:13 +0000 (14:21 +0300)]
Decouple python spec objects from transaction objects
- ts structure is only necessary for hysterically passing back the
parse result from parseSpec(), hide this in the bindings
- deprecate ts.parseSpec() for later ripping
Panu Matilainen [Wed, 23 Sep 2009 10:12:43 +0000 (13:12 +0300)]
Oops, binary data can and should be presented as python strings
Panu Matilainen [Wed, 23 Sep 2009 09:49:15 +0000 (12:49 +0300)]
Add rpmtd to python object converter, change header code to use that
- vastly simpler than the former goo in hdr_subscribe
Panu Matilainen [Wed, 23 Sep 2009 09:28:47 +0000 (12:28 +0300)]
Turn tagNumFromPyObject() into an object converter interface
- permits direct validation and conversion from arg parsing
Panu Matilainen [Wed, 23 Sep 2009 09:05:02 +0000 (12:05 +0300)]
Turn rpmFdFromPyObject() into an object converter interface
- permits direct validation and conversion from arg parsing
Panu Matilainen [Wed, 23 Sep 2009 08:51:55 +0000 (11:51 +0300)]
Permit any file object in rpm.readHeaderListFromFD()
Panu Matilainen [Wed, 23 Sep 2009 08:42:59 +0000 (11:42 +0300)]
Permit file objects in python header constructor
Panu Matilainen [Wed, 23 Sep 2009 08:41:20 +0000 (11:41 +0300)]
Add write() method to python headers
Panu Matilainen [Wed, 23 Sep 2009 08:10:38 +0000 (11:10 +0300)]
Accept any file object in ts.hdrFromFdno()
Panu Matilainen [Wed, 23 Sep 2009 08:09:19 +0000 (11:09 +0300)]
Add helper to convert python file objects to rpmio FD_t type
Panu Matilainen [Wed, 23 Sep 2009 07:49:47 +0000 (10:49 +0300)]
Add te.NEVRA() method, adjust docs
Panu Matilainen [Wed, 23 Sep 2009 07:44:41 +0000 (10:44 +0300)]
Remove tp_print methods from all rpm-python objects
- these violate the intended use of tp_print, python docs state
"A type should never implement tp_print in a way that produces
different output than tp_repr or tp_str would."
Panu Matilainen [Wed, 23 Sep 2009 07:39:40 +0000 (10:39 +0300)]
Lose the debug junk from python bindings
Panu Matilainen [Wed, 23 Sep 2009 07:31:46 +0000 (10:31 +0300)]
Add hdr.format(), deprecate hdr.sprintf()
- sprintf() is a C'ism and format() is closer to current librpm anyway
Panu Matilainen [Wed, 23 Sep 2009 07:19:46 +0000 (10:19 +0300)]
Make fiFromHeader() static inside header-py, deprecate
Panu Matilainen [Wed, 23 Sep 2009 07:11:45 +0000 (10:11 +0300)]
Deprecate rpm.headerLoad()
- this is not needed now, just call rpm.hdr(<blob) to load
Panu Matilainen [Wed, 23 Sep 2009 06:56:08 +0000 (09:56 +0300)]
Deprecate header filelist munging methods
- hdr.convert() is the generic way, the old stuff is not needed anymore
- macro to issue warnings just once
Panu Matilainen [Wed, 23 Sep 2009 06:27:47 +0000 (09:27 +0300)]
Eliminate unnecessary function
Panu Matilainen [Wed, 23 Sep 2009 06:24:23 +0000 (09:24 +0300)]
Permit headerConvert() from python
Panu Matilainen [Wed, 23 Sep 2009 06:03:25 +0000 (09:03 +0300)]
Add has_key() method to header object
Panu Matilainen [Wed, 23 Sep 2009 05:51:12 +0000 (08:51 +0300)]
Add support for creating new header objects
- without arguments, rpm.hdr() creates an empty header
- with another header as argument, make a copy
- string argument is considered an unloaded header blob and copy-loaded
Panu Matilainen [Wed, 23 Sep 2009 05:41:03 +0000 (08:41 +0300)]
Rpm's allocators dont return on OOM, remove unnecessary check
Panu Matilainen [Tue, 22 Sep 2009 21:06:13 +0000 (00:06 +0300)]
Add type checking macros for all rpm python type objects
Panu Matilainen [Tue, 22 Sep 2009 20:02:47 +0000 (23:02 +0300)]
Rename python system.h for disambiguation
Panu Matilainen [Tue, 22 Sep 2009 19:42:06 +0000 (22:42 +0300)]
Lose the empty doxygen markers
- nothing wrong with comments but empty comment placeholders
are not exactly useful
Panu Matilainen [Tue, 22 Sep 2009 18:53:21 +0000 (21:53 +0300)]
Put some consistency to python object creation
- all type object creation goes through foo_Wrap() which handle OOM
and all type specific initialization
Panu Matilainen [Tue, 22 Sep 2009 18:24:55 +0000 (21:24 +0300)]
Make all python object creation wrappers return PyObject pointers
- this way the only place where casts are needed are in the wrapper itself
Panu Matilainen [Tue, 22 Sep 2009 18:01:01 +0000 (21:01 +0300)]
Dont leak memory on rpm.ds iteration
Panu Matilainen [Tue, 22 Sep 2009 17:50:41 +0000 (20:50 +0300)]
Lose unnecessary next() methods
- python adds next() methods for objects supporting iterators
Panu Matilainen [Tue, 22 Sep 2009 17:20:07 +0000 (20:20 +0300)]
All rpm-python iterators are self-iterators, just use PyObject_SelfIter
Panu Matilainen [Tue, 22 Sep 2009 16:55:34 +0000 (19:55 +0300)]
Dont leak memory from rpm.dsSingle()
Panu Matilainen [Tue, 22 Sep 2009 16:44:59 +0000 (19:44 +0300)]
Simplify fiFromHeader()
- just call the fi object, no need to manually redo all this stuff
Panu Matilainen [Tue, 22 Sep 2009 16:42:53 +0000 (19:42 +0300)]
Simplify dsFromHeader()
- just call the ds object, no need to manually redo all this stuff
Panu Matilainen [Tue, 22 Sep 2009 16:19:02 +0000 (19:19 +0300)]
Sanitize python object -> tag number exception handling
- raise exception in tagNumFromPyObject(), not in 12 different callers
- check against RPMTAG_NOT_FOUND consistently instead of -1 and whatnot
- unknown tags are value, not key or type errors
Panu Matilainen [Tue, 22 Sep 2009 15:55:09 +0000 (18:55 +0300)]
Move allocations out of rpmps object init method
- tp_init can be called several times, allocating from there leaks memory
- tp_init gets called automatically on object creation, dont call manually
- nothing to initialize for rpmps...
Panu Matilainen [Tue, 22 Sep 2009 15:53:15 +0000 (18:53 +0300)]
Move allocations out of rpmds object init method
- tp_init can be called several times, allocating from there leaks memory
- tp_init gets called automatically on object creation, dont call manually
Panu Matilainen [Tue, 22 Sep 2009 15:50:32 +0000 (18:50 +0300)]
Move allocations out of rpmfi object init method
- tp_init can be called several times, allocating from there leaks memory
- tp_init gets called automatically on object creation, dont call manually
Panu Matilainen [Tue, 22 Sep 2009 15:22:50 +0000 (18:22 +0300)]
Remove useless rpmts python initialization
Panu Matilainen [Tue, 22 Sep 2009 15:17:49 +0000 (18:17 +0300)]
Realize the remaining bits of direct rpmdb interface are dead too
- the code hasn't been enabled in a long long time despite being built...
Panu Matilainen [Tue, 22 Sep 2009 14:49:53 +0000 (17:49 +0300)]
Use Py_RETURN_NONE macro for returning None everywhere
Panu Matilainen [Tue, 22 Sep 2009 14:33:07 +0000 (17:33 +0300)]
Eliminate unnecessary custom object allocation functions
- these are just calling python defaults, no point at all...
Panu Matilainen [Tue, 22 Sep 2009 14:27:44 +0000 (17:27 +0300)]
Use generic python get/set attribute functions directly where appropriate
- no point in wrapping all this stuff in our own functions
Panu Matilainen [Tue, 22 Sep 2009 14:12:22 +0000 (17:12 +0300)]
Python initialization spring clean
- split rpmtag enum additions to seprate function
- PyModule_AddIntConstant() does exactly what we need for the rest
Panu Matilainen [Tue, 22 Sep 2009 13:47:07 +0000 (16:47 +0300)]
Make the python object structures opaque
Panu Matilainen [Tue, 22 Sep 2009 13:39:39 +0000 (16:39 +0300)]
Lose the utterly broken rpmio FD_t wrappers
Panu Matilainen [Tue, 22 Sep 2009 13:36:48 +0000 (16:36 +0300)]
Lose the useless python debug stuff
Panu Matilainen [Tue, 22 Sep 2009 13:28:58 +0000 (16:28 +0300)]
We dont support ancient python versions...
Panu Matilainen [Tue, 22 Sep 2009 13:24:32 +0000 (16:24 +0300)]
Kick out mummified rpmdb python interface remnants
Panu Matilainen [Tue, 22 Sep 2009 13:02:39 +0000 (16:02 +0300)]
Missing <string.h> include
Panu Matilainen [Tue, 22 Sep 2009 12:36:11 +0000 (15:36 +0300)]
Clean up + fix bad recno erasure in python ts.addErase()
- now that we can loop over RPMDBI_PACKAGES matches, the code can
be simplified quite a bit
- walk through the iterator always to get a reliable installed status,
previously ts.addErase() with bad record number wouldn't raise exception
Panu Matilainen [Tue, 22 Sep 2009 11:53:56 +0000 (14:53 +0300)]
Remove no longer necessary RPMDBI_PACKAGES special casing
Panu Matilainen [Tue, 22 Sep 2009 11:52:57 +0000 (14:52 +0300)]
Fix endless loop on RPMDBI_PACKAGES iteration when offset specified
Panu Matilainen [Tue, 22 Sep 2009 09:32:02 +0000 (12:32 +0300)]
Unbreak --tid and --querybynumber
- broken ages ago somewhere between 4.4.x and 4.6.x by typing changes
presumably
- C99 strtoul() can figure the base on its own, remove unnecessary work
Panu Matilainen [Tue, 22 Sep 2009 07:21:47 +0000 (10:21 +0300)]
DB_INCOMPLETE was removed in BDB 4.1 and we dont support older versions
Panu Matilainen [Tue, 22 Sep 2009 06:26:39 +0000 (09:26 +0300)]
Remove ancient, unused code to filter duplicates on rpmdb rebuild
- enabling this as-is would be murderous to multilib systems
- getting duplicates in the db would be a bug elsewhere
Panu Matilainen [Mon, 21 Sep 2009 11:36:59 +0000 (14:36 +0300)]
Rip out dependency caching
- it doesn't speed up things that much, is broken in some chroot
scenarios and is ugly ugly hardwired BDB hackery where it doesn't belong
Panu Matilainen [Mon, 21 Sep 2009 11:04:34 +0000 (14:04 +0300)]
Move rpmdb home creation from backends to one level higher
- this is a common operation for any backends, avoid duplication
- only try to create it once instead of every index open, and if it
doesn't exist and cant be created, error out instead of trying to continue
in vain
Panu Matilainen [Mon, 21 Sep 2009 11:00:46 +0000 (14:00 +0300)]
Add an internal API for getting rpmdb home wrt chroot state, use it
Panu Matilainen [Mon, 21 Sep 2009 08:59:25 +0000 (11:59 +0300)]
Remove pointless per-index api version
Panu Matilainen [Mon, 21 Sep 2009 08:36:16 +0000 (11:36 +0300)]
DB config stuff not needed outside dbconfig.c, make them static
- additionally make the popt struct const since it can...
Panu Matilainen [Mon, 21 Sep 2009 08:32:43 +0000 (11:32 +0300)]
BDB is not optional at least currently, remove useless ifdefs
Panu Matilainen [Mon, 21 Sep 2009 08:29:27 +0000 (11:29 +0300)]
Remove useless no-op "teardown" and "usecursors" config bits
Panu Matilainen [Mon, 21 Sep 2009 08:00:50 +0000 (11:00 +0300)]
Remove BDB RPC "support"
- it's insecure, used by nobody (its been rather broken for a long
time) and all, but BDB 4.8.x removed RPC support from upstream so we
cant very well support in the future even if we wanted to
Panu Matilainen [Mon, 21 Sep 2009 07:29:38 +0000 (10:29 +0300)]
Remove XA "support" from the backend
- it's removed from BDB 4.8.x and this has hardly ever done anything anyway
Panu Matilainen [Fri, 18 Sep 2009 10:53:46 +0000 (13:53 +0300)]
Remove unused db/dbenv methods "documentary" from the source
Panu Matilainen [Fri, 18 Sep 2009 10:05:49 +0000 (13:05 +0300)]
Use db3Free() for on sqlite backend close too
Panu Matilainen [Fri, 18 Sep 2009 10:13:59 +0000 (13:13 +0300)]
Always populate dbi->dbi_file on db3New()
- Only the actual db->open() wants NULL for temporary databases, everywhere
else we want the tag name. Handle the temporary case where it counts
instead of complicating things everywhere for the one special case.
Panu Matilainen [Fri, 18 Sep 2009 08:56:05 +0000 (11:56 +0300)]
Mop up leftover root/home foobar
Panu Matilainen [Fri, 18 Sep 2009 08:51:21 +0000 (11:51 +0300)]
Per-dbi root and home dont make sense, always use rpmdb paths
Panu Matilainen [Fri, 18 Sep 2009 06:49:28 +0000 (09:49 +0300)]
Sigh.. when not using environment, absolute paths are required
Panu Matilainen [Thu, 17 Sep 2009 13:28:09 +0000 (16:28 +0300)]
Always open database files relative to database home
- This allows the db to move freely wherever (such as across chroots,
RhBug:507309 and pile of related issues) .. and is even recommended
by BDB documentation. Well duh.
Panu Matilainen [Thu, 17 Sep 2009 09:51:28 +0000 (12:51 +0300)]
EXIT_FAILURE being defined is already handled in system.h
Panu Matilainen [Thu, 17 Sep 2009 09:40:29 +0000 (12:40 +0300)]
Compare pointers against NULL, not 0
Panu Matilainen [Thu, 17 Sep 2009 09:25:31 +0000 (12:25 +0300)]
Have BDB use our allocation routines now that we can
- rfree() prototype doesn't match free() but it doesn't matter in this
case as it's just a convencience wrapper with no extra functionality
and setting NULL is ok according to BDB docs
- db->set_alloc() can't be used if environment is used, fix the condition
Panu Matilainen [Thu, 17 Sep 2009 08:36:10 +0000 (11:36 +0300)]
Add runtime settable custom memory allocation failure callback routine
- lets API users perform theirn own cleanup / exit through their own
routines in case of allocation failure, or even try to free up
some memory and retry allocation
Panu Matilainen [Wed, 16 Sep 2009 14:28:22 +0000 (17:28 +0300)]
Always use build + use our own allocator functions
- Consistent across platforms and will allow some things macros dont and
memory checkers these days are smart enough to get decent output anyway
- Minimal namespacing with r-prefix, add compatibility macros to system.h
for now so we dont have to change the entire codebase for this
- Also make rpmutil.h where the declarations and gcc __attribute__ macros
are available everywhere
Panu Matilainen [Wed, 16 Sep 2009 13:18:06 +0000 (16:18 +0300)]
Eliminate unnecessary variable
Florian Festi [Wed, 16 Sep 2009 12:59:48 +0000 (14:59 +0200)]
Remove unused local variable
Panu Matilainen [Wed, 16 Sep 2009 12:36:44 +0000 (15:36 +0300)]
Remove db environment on close on chrooted operations
- As we open the db from outside the chroot, the environment ends up
containing paths that are not valid once we enter the chroot causing
dumb issues like RhBug:513699 and RhBug:507309.
- We'd be better off removing the environment always after access but
limiting to chroots for now as the rpmdb open/close path is full of races
when environment is not present. Chroots are somewhat special environemnts
anyway and typically not concurrently accessed so it's less of an issue
there.
- While this still has all sorts of issues, it at least leaves the rpmdb
in chroot in a functional state after initial install.
Panu Matilainen [Wed, 16 Sep 2009 12:12:28 +0000 (15:12 +0300)]
Move the default transaction lock out of BDB environment namespace
- We dont want the transaction lock to get mixed with the BDB environment,
nor do we want it wiped out with it as the ts lock protects things
*around* the environment. Unlike the environment, any fcntl locks on
the file get released when process dies, so if there is a lock,
there's a process alive holding it and you'd better not remove it.
- Use the same lock path in macro configuration and built-in fallback
- Make it a hidden file to avoid people confusing lock file existence
with it actually being locked
Panu Matilainen [Wed, 16 Sep 2009 11:08:28 +0000 (14:08 +0300)]
There's a function to test string equivalence...
Panu Matilainen [Wed, 16 Sep 2009 11:00:54 +0000 (14:00 +0300)]
Throw out useless db statistics structs
- besides being completely unused, these structs are defined by BDB
public headers, no need to mirror them in rpm
Panu Matilainen [Wed, 16 Sep 2009 10:57:01 +0000 (13:57 +0300)]
Throw out pile of unused "stuff" from 2002
Panu Matilainen [Wed, 16 Sep 2009 09:41:44 +0000 (12:41 +0300)]
Minor cleanup to newRpmdb()
- no need to free what hasn't been allocated yet
Panu Matilainen [Wed, 16 Sep 2009 09:27:12 +0000 (12:27 +0300)]
One complicated way to write xstrdup() .. pff
- sqlite backend uses dbi subfile for table name, leave it alone although
it is the same as the file name...
- allocate + free both file and table name separately
Panu Matilainen [Wed, 16 Sep 2009 09:08:58 +0000 (12:08 +0300)]
Eliminate long since unused db subfile from bdb code
Panu Matilainen [Wed, 16 Sep 2009 08:56:12 +0000 (11:56 +0300)]
Remove unnecessary arguments to db init and fini
- bdb environments are per db_home, not per file so dbfile etc dont make
sense here
Florian Festi [Mon, 14 Sep 2009 14:19:33 +0000 (16:19 +0200)]
Also skip to be erased files in "netsharedpath"
fixes rhbz#494640
Panu Matilainen [Tue, 15 Sep 2009 09:57:51 +0000 (12:57 +0300)]
Initialize rpm earlier in rpm2cpio (RhBug:523260)
- url retrieval requires macros to be loaded
Panu Matilainen [Tue, 15 Sep 2009 08:05:27 +0000 (11:05 +0300)]
Take non-gi errors into account too (SuseBug:527191)
- patch from OpenSUSE / Michael Schroeder