Panu Matilainen [Wed, 2 Jul 2008 10:31:12 +0000 (13:31 +0300)]
Some keyring improvements
- keep keyring sorted by keyid, use bsearch for key lookups
- check for key in keyring before adding it to avoid dupes
- actually check that pubkey parameters match the signature on keyring
lookup like was done in rpmtsFindPubkey()
Panu Matilainen [Tue, 1 Jul 2008 14:35:21 +0000 (17:35 +0300)]
Eliminate remaining signature checking bits from rpmts
- rpmdb-specific rpmtsFindPubkey() no longer needed for anything, and
neither are the cached packets and stuff
Panu Matilainen [Tue, 1 Jul 2008 14:32:11 +0000 (17:32 +0300)]
Use keyring even when using pubkeys from rpmdb
- build-time switchable, could easily be run-time too
Panu Matilainen [Tue, 1 Jul 2008 14:00:04 +0000 (17:00 +0300)]
Pass keyring, not the full ts to lowlevel signature routines
- lazy loading of default keyring on ts iteration works nicely for
everything but package signature verification which doesn't access the
db now that the keys are not in the db anymore
- make rpmtsSetKeyring(NULL) load the default keyring, a bit ugly but
it'll have to do for now
Panu Matilainen [Tue, 1 Jul 2008 12:50:05 +0000 (15:50 +0300)]
Eliminate digest calculation statistics from signature code
- if we want them back, they just need to be detached from the ts...
Panu Matilainen [Tue, 1 Jul 2008 12:46:18 +0000 (15:46 +0300)]
Add methods for getting and setting the keyring used by a transaction
- semantics of exactly when and how this is permitted needs more thought...
Panu Matilainen [Tue, 1 Jul 2008 12:38:32 +0000 (15:38 +0300)]
Move keyring related typedefs to rpmtypes.h
Panu Matilainen [Tue, 1 Jul 2008 12:25:06 +0000 (15:25 +0300)]
Proof of concept rpmtsFindPubkey() implementation with the new keyring
- ifdeffed out for now, needs a bit more flesh and a working
import routine
Panu Matilainen [Tue, 1 Jul 2008 12:24:04 +0000 (15:24 +0300)]
Beginnings of an abstract keyring
- in librpmio to avoid dependencies on higher level stuff
- for now it's just an in-memory structure that needs to be populated
by someone simply by feeding in GPG public keys
Panu Matilainen [Tue, 1 Jul 2008 11:35:44 +0000 (14:35 +0300)]
Create ascii armor that rpm itself accepts
Jindrich Novy [Tue, 1 Jul 2008 10:52:16 +0000 (12:52 +0200)]
Update parseNoSource() and addSource() to accept only unsigned SourceXX numbers
Jindrich Novy [Tue, 1 Jul 2008 10:51:08 +0000 (12:51 +0200)]
findSource() eats only unsigned SourceXX numbers
Jindrich Novy [Tue, 1 Jul 2008 10:49:19 +0000 (12:49 +0200)]
doPatchMacro() now uses parseUnsignedNum()
Jindrich Novy [Tue, 1 Jul 2008 10:48:50 +0000 (12:48 +0200)]
doSetupMacro() now uses parseUnsignedNum()
Jindrich Novy [Tue, 1 Jul 2008 10:47:36 +0000 (12:47 +0200)]
doUntar() accepts only unsigned SourceXX numbers
Jindrich Novy [Tue, 1 Jul 2008 10:46:16 +0000 (12:46 +0200)]
Convert doPatch() to use parseUnsignedNum()
Jindrich Novy [Tue, 1 Jul 2008 10:43:47 +0000 (12:43 +0200)]
Use the same number types for SourceXX tags as returned by parseUnsignedNum()
Jindrich Novy [Tue, 1 Jul 2008 10:42:12 +0000 (12:42 +0200)]
Convert Epoch parsing to use parseUnsignedNum()
Jindrich Novy [Tue, 1 Jul 2008 10:41:35 +0000 (12:41 +0200)]
Avoid reading negative constants from SourceXX, PatchXX tags
- convert parseNum() to parseUnsignedNum()
- don't misleadingly declare it to return int, when the result is
always unsigned due to strtoul() call
Panu Matilainen [Tue, 1 Jul 2008 08:22:40 +0000 (11:22 +0300)]
Eliminate the nasty global _dig and _digp variables from rpmpgp
- pass the digest + parameters around as arguments instead
Panu Matilainen [Mon, 16 Jun 2008 14:49:02 +0000 (17:49 +0300)]
Unexport all the umphteen pgp related tables
- nothing outside rpmpgp.c needs
Panu Matilainen [Mon, 16 Jun 2008 14:01:29 +0000 (17:01 +0300)]
Make every pgp function not needed by other parts of rpm static
Panu Matilainen [Mon, 16 Jun 2008 08:12:15 +0000 (11:12 +0300)]
We don't need ts to stash the keyid locally now
Panu Matilainen [Mon, 16 Jun 2008 07:59:23 +0000 (10:59 +0300)]
headerCheck() does not recurse
- it can be re-entered by the way of rpmtsFindPubkey() but that doesn't
matter now as dig etc are local variables
Panu Matilainen [Wed, 25 Jun 2008 08:17:17 +0000 (11:17 +0300)]
Detach pgp digest from transaction set
- pass the pgp container around as argument as needed
- eliminate the related API from rpmts
Panu Matilainen [Sat, 14 Jun 2008 11:48:53 +0000 (14:48 +0300)]
Pass pgpDig around from rpmVerifySignature()
- get the digest once in rpmVerifySignature() and pass to verification
functions, signature is part of the digest so no need for separate call..
Panu Matilainen [Sat, 14 Jun 2008 14:09:50 +0000 (17:09 +0300)]
Detach signature tag data from transaction set
- pass signature tag data around as argument to rpmVerifySignature() as
needed
- eliminate the related API from rpmts
Panu Matilainen [Sat, 14 Jun 2008 11:03:11 +0000 (14:03 +0300)]
Actually use the sigtd passed to verifier functions
Panu Matilainen [Sat, 14 Jun 2008 10:49:37 +0000 (13:49 +0300)]
Start getting signature tag out of transaction set
- collect various signature items into rpmtd and pass around to
verifyFooSignature() checkers
Panu Matilainen [Mon, 30 Jun 2008 12:13:54 +0000 (15:13 +0300)]
(char)NULL doesn't make much sense, use \0 instead
- shuts up another gcc warning
Jindrich Novy [Mon, 30 Jun 2008 10:23:42 +0000 (12:23 +0200)]
Fix missspelling in macro name (_initddir -> _initrddir), thanks to
- Ville Skyttä
- old macro with typo is kept for compatiblity
Panu Matilainen [Mon, 30 Jun 2008 08:28:32 +0000 (11:28 +0300)]
size_t for "left" is wrong, use rpm_loff_t instead
Panu Matilainen [Mon, 30 Jun 2008 08:24:34 +0000 (11:24 +0300)]
strntoul() is supposed to return unsigned long, not int
- the type mismatch busts up handling on largish files...
Panu Matilainen [Fri, 27 Jun 2008 12:15:46 +0000 (15:15 +0300)]
debugedit: Include empty CU current directories (rhbz#444310)
- patch from Jan Kratochvil
Panu Matilainen [Fri, 27 Jun 2008 11:39:39 +0000 (14:39 +0300)]
Use correct formatters for debug-printing in rpmtsInitDSI()
Panu Matilainen [Fri, 27 Jun 2008 10:49:48 +0000 (13:49 +0300)]
Use correct formatters for debug-printing sizes in printSize()
- on x86_64 size_t happens to be 64bit, not so on x86...
Panu Matilainen [Fri, 27 Jun 2008 10:02:23 +0000 (13:02 +0300)]
Adjust python callback for amount+total type change
Panu Matilainen [Thu, 26 Jun 2008 14:52:31 +0000 (17:52 +0300)]
Argh, the tag is LONGSIZE, not LARGESIZE
Jindrich Novy [Thu, 26 Jun 2008 13:58:40 +0000 (15:58 +0200)]
Take new directory into account in rpmbuild test
Jindrich Novy [Thu, 26 Jun 2008 13:56:55 +0000 (15:56 +0200)]
Add new BUILDROOT directory to default rpm build dirs.
Jindrich Novy [Thu, 26 Jun 2008 13:56:11 +0000 (15:56 +0200)]
Add new macros for enabling default build root
Panu Matilainen [Thu, 26 Jun 2008 13:37:06 +0000 (16:37 +0300)]
Make --info work on large packages too
Panu Matilainen [Thu, 26 Jun 2008 13:34:32 +0000 (16:34 +0300)]
Teach build about 64bit payload etc size tags
- only use 64bit variants for packages that actually need it to avoid
breaking compatibility when not needed
- old rpm will bail out "cleanly" (with an obscure error message) if
the 32bit size tags are missing but at least it doesn't crash...
Panu Matilainen [Thu, 26 Jun 2008 13:28:17 +0000 (16:28 +0300)]
Teach rpmAddSignature() and rpmReadSignature() about 64bit tags
- use 64bit size if it exists, otherwise quietly fall back to 32bit one
- can't use tag extensions to convert there as legacy sigtags conflict
with things like RPMTAG_NAME
Panu Matilainen [Thu, 26 Jun 2008 13:23:56 +0000 (16:23 +0300)]
RPMTAG_LONGSIGSIZE and RPMTAG_LONGARCHIVESIZE are signature tags
- move them from regular tag space to sigtag space, fortunately these
don't require special conversion
Panu Matilainen [Thu, 26 Jun 2008 13:10:49 +0000 (16:10 +0300)]
Only insert 64bit filesizes on build if required
- older rpm's get pretty upset on encounter with 64bit integers, avoid
unnecessary breakage as very few packages actually need large file sizes
- Add per-filelist flag for large files, enabled from addFile() as we
check sizes. It's basically a no-op for now as we error out if any
single file exceeds cpio per file limit.
Panu Matilainen [Thu, 26 Jun 2008 10:59:06 +0000 (13:59 +0300)]
Unbreak compressFilelist()
- well and truly broken by yours truly by a Monday morning commit
efe1665024e374a8f8704d229b1d306af232c396, blush...
- can't use headerIsSource() here, as build calls compressFilelist()
before RPMTAG_SOURCERPM is added to the header, revert to earlier hack
and check filename beginning with / or not
- dncmp() expects address of a string, not the string itself, this was
masked by the above bug
Panu Matilainen [Thu, 26 Jun 2008 08:38:28 +0000 (11:38 +0300)]
Add a bit of protection against oversized files
- max size of individual files in cpio new ascii format is well known,
check we don't try to stuff in larger than what fits and abort cleanly
- this should do as long as we only support new ascii format cpio payloads
Panu Matilainen [Thu, 26 Jun 2008 07:50:01 +0000 (10:50 +0300)]
Type corrections
- Fread() and Fwrite() return ssize_t, not int32_t
- sigtag is rpmSigTag, not int32_t
Panu Matilainen [Wed, 25 Jun 2008 07:59:20 +0000 (10:59 +0300)]
Fix handling of "%posttrans -p /xxx" with no body (Pascal Rigaux)
Panu Matilainen [Wed, 25 Jun 2008 06:56:12 +0000 (09:56 +0300)]
Adjust autotest querytag expectations to match current tags
Panu Matilainen [Wed, 25 Jun 2008 06:53:21 +0000 (09:53 +0300)]
Adjust autotests for rpmi etc removal
Panu Matilainen [Wed, 25 Jun 2008 06:33:50 +0000 (09:33 +0300)]
Force "rpm" popt context for everything but build
- we want the same set of aliases to be seen no matter what name we're
invoked with, except for build which has it's own set of aliases
- this has a nice little side-effect of permitting popt aliases to work
from source tree on non-static builds despite libtool calling executables
lt-something instead of their real names
Panu Matilainen [Wed, 25 Jun 2008 06:20:27 +0000 (09:20 +0300)]
--buildpolicy is rpmbuild, not rpm, alias
Panu Matilainen [Wed, 25 Jun 2008 06:19:13 +0000 (09:19 +0300)]
Kill the million duplicate aliases in rpmpopt
Panu Matilainen [Wed, 25 Jun 2008 06:18:51 +0000 (09:18 +0300)]
Eliminate rpmd, rpmi, rpmk, rpmq "helper binaries"
- it's all rolled into "rpm" itself, these are just clutter and complicate
things needlessly
- leave symlinks to rpmquery, rpmverify, rpmsign and rpmdb -> rpm to
preserve some compatibility (at least rpmquery is known to be used
in scripts)
Panu Matilainen [Tue, 24 Jun 2008 13:07:05 +0000 (16:07 +0300)]
Update rpm manpage wrt build modes
- the build modes have been removed ages ago, just point to rpmbuild manual
Panu Matilainen [Tue, 24 Jun 2008 12:23:30 +0000 (15:23 +0300)]
Eliminate build-related aliases on "rpm"
- "rpm" for building packages hasn't worked in ~six years, time to clean
up the junk from rpmpopt
Panu Matilainen [Tue, 24 Jun 2008 08:11:38 +0000 (11:11 +0300)]
Eliminate already commented out build mode aliases for rpm
- these haven't been enabled in six years and not coming back...
Panu Matilainen [Tue, 24 Jun 2008 08:09:00 +0000 (11:09 +0300)]
Eliminate rpmb aliases from rpmpopt
Panu Matilainen [Tue, 24 Jun 2008 08:08:05 +0000 (11:08 +0300)]
Eliminate the pointless rpmb + rpmt symlinkery
- just build "rpmbuild" executable and install to bindir instead of
mucking with several different names
Panu Matilainen [Tue, 24 Jun 2008 07:08:37 +0000 (10:08 +0300)]
Force FD_CLOEXEC on all potentially open descriptors
- instead of just "100 should be large enough", use sysconf() to grab
number of max open files and do them all. If sysconf() fails,
use 1024 as "should be enough for everybody"
Panu Matilainen [Mon, 23 Jun 2008 13:57:51 +0000 (16:57 +0300)]
Remove artificial limit in dependency loop elimination attempts
- continue processing as long as progress can be made instead of artificial
hardcoded magic "try ten times"
Panu Matilainen [Mon, 23 Jun 2008 12:02:10 +0000 (15:02 +0300)]
Avoid silly deprecation warning on headerFreeData()
Panu Matilainen [Mon, 23 Jun 2008 11:37:54 +0000 (14:37 +0300)]
Simplify requires and provides addition in rpmfcGenerateDepends()
- just feed (name, evr and flags) one by one into header instead of
constructing arrays of them
Panu Matilainen [Mon, 23 Jun 2008 11:31:54 +0000 (14:31 +0300)]
Oops, provideflags were being inserted as requireflags
Panu Matilainen [Mon, 23 Jun 2008 05:46:51 +0000 (08:46 +0300)]
Add Malay translation from Sharuzzaman Ahmat Raslan
Panu Matilainen [Thu, 19 Jun 2008 13:08:00 +0000 (16:08 +0300)]
Dumb, dumb thinko/leftover from earlier prototype...
Panu Matilainen [Thu, 19 Jun 2008 12:52:54 +0000 (15:52 +0300)]
Use type-specific headerPuts() for rpmcache
- ...to finish of undoing yesterdays damage, sigh ;)
Panu Matilainen [Thu, 19 Jun 2008 12:52:19 +0000 (15:52 +0300)]
Use type-specific headerPut() for the case where it makes sense in relocs
Panu Matilainen [Thu, 19 Jun 2008 12:51:07 +0000 (15:51 +0300)]
Use type-specific headerPut() in rpmpsmStage()
Panu Matilainen [Thu, 19 Jun 2008 12:50:32 +0000 (15:50 +0300)]
Use type-specific headerPut() for installtid stuff in rpmdbAdd()
Panu Matilainen [Thu, 19 Jun 2008 12:49:09 +0000 (15:49 +0300)]
Use type-specific headerPut()'s for rpmtsImportPubkey()
Panu Matilainen [Thu, 19 Jun 2008 12:48:31 +0000 (15:48 +0300)]
Use type-specific headerPut()'s for legacy retrofitting stuff
Panu Matilainen [Thu, 19 Jun 2008 12:47:04 +0000 (15:47 +0300)]
Use type-specific headerPut() where makes sense in build/parseScript
Panu Matilainen [Thu, 19 Jun 2008 12:46:34 +0000 (15:46 +0300)]
Use type-specific headerPut() where makes sense in build/rpmfc.c
Panu Matilainen [Thu, 19 Jun 2008 12:45:57 +0000 (15:45 +0300)]
use type-specific headerPut() on all of build/reqprov.c
Panu Matilainen [Thu, 19 Jun 2008 12:45:37 +0000 (15:45 +0300)]
Use type-specific headerPut() on all of build/parseSpec.c
Panu Matilainen [Thu, 19 Jun 2008 12:45:06 +0000 (15:45 +0300)]
Use type-specific headerPut() for all of build/parsePreamble.c
Panu Matilainen [Thu, 19 Jun 2008 12:44:29 +0000 (15:44 +0300)]
Use type-specific headerPut() for all of build/parseChangelog.c
Panu Matilainen [Thu, 19 Jun 2008 12:43:45 +0000 (15:43 +0300)]
Use type specific headerPut() for all of build/pack.c
Panu Matilainen [Thu, 19 Jun 2008 12:43:14 +0000 (15:43 +0300)]
Use type specific headerPut() for all of build/files.c
Panu Matilainen [Thu, 19 Jun 2008 12:41:58 +0000 (15:41 +0300)]
Add type-safe headerPutFoo() methods for supported tag data types
- requiring all access through rpmtdFromFoo() just adds unnecessary
indirection and pain for little gain, the header is fairly intelligent
when it comes to inserting data
- this gives us various benefits over the old headerAddEntry() interface too:
+ basic type checking done by compiler
+ extra sanity checking (ie you can't add string data to integer tags,
don't permit adding more than one entry to non-array data etc)
+ "do the right thing" approach - add / append as needed and supported
by various types
+ headerPutString() can now be used on both single strings and string
arrays making the interface much nicer for the rather common case of
appending strings one by one to string array tags
Panu Matilainen [Wed, 18 Jun 2008 13:14:36 +0000 (16:14 +0300)]
Dumb thinko breaking scriptlet interpreter
Panu Matilainen [Wed, 18 Jun 2008 12:39:59 +0000 (15:39 +0300)]
Group legacy interfaces at the end of header.h
- out of sight, out of mind ;)
- add deprecation warnings where missing
- move to header_legacy doxygen group to avoid littering main documentation
Panu Matilainen [Wed, 18 Jun 2008 12:22:46 +0000 (15:22 +0300)]
Convert headerAddI18NString() to use headerPut()
Panu Matilainen [Wed, 18 Jun 2008 12:16:05 +0000 (15:16 +0300)]
Switch around headerAdd*Entry() family vs headerPut()
- make internal header add + append take rpmtd as argument, legacy
interfaces are just wrappers around them
- add deprecation markers
Panu Matilainen [Wed, 18 Jun 2008 11:59:13 +0000 (14:59 +0300)]
Switch around headerMod() and headerModifyEntry()
- headerModifyEntry() is the legacy interface going away, make it
use headerMod() and deprecate it
Panu Matilainen [Wed, 18 Jun 2008 11:54:41 +0000 (14:54 +0300)]
Stuff legacy compat wrappers out of sight into header_internal.c
- probably need to carry the compat stuff around for a while but get
them out of the way...
- headerFreeTag() and headerFreeData() are just the same thing, with
headerFreeTag() supposedly being the "new" interface which never was
used, duh :)
Panu Matilainen [Wed, 18 Jun 2008 11:35:56 +0000 (14:35 +0300)]
Start shredding legacy cruft out of header code
- headerGetRaw() can now be done with just a flag to headerGet()
Panu Matilainen [Wed, 18 Jun 2008 10:15:46 +0000 (13:15 +0300)]
Convert signature code to use headerPut()
- tags in signature header overlap with those in "regular" header,
can't use rpmtdFromFoo() as that would give totally bogus types and errors
- just add a wrapper around headerPut() that behaves like headerAddEntry()
Panu Matilainen [Wed, 18 Jun 2008 10:13:22 +0000 (13:13 +0300)]
Convert parseScript() to headerPut()
- AFAIR the only place in rpm where the type of a tag varies like this, ugh
- any modern rpm can handle the type being either string or string array,
but preserving legacy behavior for now...
Panu Matilainen [Wed, 18 Jun 2008 09:57:36 +0000 (12:57 +0300)]
Convert rpmpsmStage() to use headerPut()
Panu Matilainen [Wed, 18 Jun 2008 09:55:02 +0000 (12:55 +0300)]
Convert rpmdbAdd() to use headerPut()
Panu Matilainen [Wed, 18 Jun 2008 09:52:28 +0000 (12:52 +0300)]
Convert rpmtsImportPubkey() to use headerPut()
Panu Matilainen [Wed, 18 Jun 2008 09:37:42 +0000 (12:37 +0300)]
Convert RPM_BIN_TYPE uses in build to headerPut()
Panu Matilainen [Wed, 18 Jun 2008 09:36:20 +0000 (12:36 +0300)]
Add rpmtdFromUint8()
- handles CHAR, INT8 and BIN types
Panu Matilainen [Wed, 18 Jun 2008 08:32:59 +0000 (11:32 +0300)]
Convert legacy retrofits & friends to headerPut()
Panu Matilainen [Wed, 18 Jun 2008 08:16:04 +0000 (11:16 +0300)]
Lose the remaining hae and hme vectors from rpmfi
Panu Matilainen [Wed, 18 Jun 2008 08:13:58 +0000 (11:13 +0300)]
Convert file relocation to use headerPut/Mod