Panu Matilainen [Mon, 14 Apr 2008 11:27:29 +0000 (14:27 +0300)]
Don't bother calculating max prefix lengths
- no longer needed, prefixBuf is dynamically allocated as needed
Jindrich Novy [Mon, 14 Apr 2008 11:07:58 +0000 (13:07 +0200)]
Prevent buffer overflow in readKeys()
Jindrich Novy [Fri, 11 Apr 2008 13:52:56 +0000 (15:52 +0200)]
Use dynamic allocation in rpmlibNeedsFeature()
Jindrich Novy [Fri, 11 Apr 2008 13:44:04 +0000 (15:44 +0200)]
Don't warn about BuildRoot being "/" while it is actually empty
Panu Matilainen [Fri, 11 Apr 2008 08:45:35 +0000 (11:45 +0300)]
Simplify url retrieval in rpmInstall()
- use rpmMkTemp() instead of manual kludgery to static buffer
Panu Matilainen [Fri, 11 Apr 2008 08:18:25 +0000 (11:18 +0300)]
Simplify url io a lot
- move urlhelper call to urlGetFile()
- have urlOpen() call urlGetFile() instead of going the other way around
and causing local copies and all sorts of craziness in the process
Panu Matilainen [Fri, 11 Apr 2008 06:38:44 +0000 (09:38 +0300)]
Remove #ifdef NOTYET junk
Panu Matilainen [Fri, 11 Apr 2008 06:05:05 +0000 (09:05 +0300)]
Rewrite rpmMkTempFile() for sanity
- Actually use mkstemp() for creating the temp file and return a FD_t
dupped from the file descriptor returned by mkstemp().
- Simplify the interface while at it.
- Change callers for the new interface.
- Yes we now require mkstemp() to work, mkstemp() is in POSIX.1-2001 and
this is year 2008...
Panu Matilainen [Thu, 10 Apr 2008 16:36:03 +0000 (19:36 +0300)]
First crack at adding ISA provides to packages (rhbz#235755)
- Horrible kludgery to get the isa names and bits into platform specific
macros from installplatform script. That beast needs to die. I mean really
- In build, add provides: name(isa) = evr automatically when it makes
sense (similarly to name = evr provides). ISA consists of ISA name and
bitness (or wordsize). This can be used to correctly
express multilib dependencies without resorting to (expensive!) file
dependency kludges, eg for dlopen()'ed libraries where automatic
dep extraction doesn't force dependency on 32bit vs 64bit version, you
can now use:
Requires: foo-plugin%{?_isa}
This expands to foo-plugin(x86-32) for i?86 packages, foo-plugin(x86-64)
to x86_64 etc, and permits spec to be shared with older distros which
don't have ISA provides.
- The same could be expressed with "canon arch" just as well, but
using the ISA to differentiate from %_arch and the like:
eg i386 could be used instead of x86-32 but it's overloaded with meanings
(the actual i386 processor vs i386 compatible cpu family etc)
Panu Matilainen [Thu, 10 Apr 2008 13:07:47 +0000 (16:07 +0300)]
Avoid exporting providePackageNVR() in the library
Panu Matilainen [Thu, 10 Apr 2008 13:04:12 +0000 (16:04 +0300)]
Just call legacyRetrofit() instead of combined compress + provide retrofits
Panu Matilainen [Thu, 10 Apr 2008 12:52:36 +0000 (15:52 +0300)]
Add a static, far simpler addPackageProvides() to build
- use instead of providePackageNVR() which needs to deal with all sorts
of hysteric "rpm 3.0.2 did this but 3.0.5 didn't" issues
Jindrich Novy [Thu, 10 Apr 2008 12:42:50 +0000 (14:42 +0200)]
Call rpmReadConfigFiles() in rpm2cpio so that rpm2cpio won't complain like:
warning: Unknown system: (null)
warning: Please contact rpm-maint@lists.rpm.org
Panu Matilainen [Thu, 10 Apr 2008 08:25:59 +0000 (11:25 +0300)]
Remove unnecessary providePackageNVR() from writeRPM()
- packageBinaries() already calls providePackageNVR(), no need to do it twice
Panu Matilainen [Thu, 10 Apr 2008 08:19:40 +0000 (11:19 +0300)]
Unbreak srpm generation broken by
6c4b0cbd93709a0fa18af1ef07af44da585bb5f6
- cpio path in source packages is the basename of the file, make this
obvious by using basename() instead of pointer games
Jindrich Novy [Wed, 9 Apr 2008 11:05:45 +0000 (13:05 +0200)]
Don't use stack allocations in rpm2cpio when not needed
Panu Matilainen [Wed, 9 Apr 2008 10:54:39 +0000 (13:54 +0300)]
Replace alloca+memset with xcalloc() in rpmdbRemove()
- return value is suspect: should return "ret" instead of 0 at exit
depending on what happened in removal but preserving previous behavior
for now...
Panu Matilainen [Wed, 9 Apr 2008 10:48:00 +0000 (13:48 +0300)]
Replace alloca+memset with xcalloc() in rpmdbAdd()
Panu Matilainen [Wed, 9 Apr 2008 10:47:29 +0000 (13:47 +0300)]
Replace alloca+memset with xcalloc() in rpmdbCountPackages()
Panu Matilainen [Wed, 9 Apr 2008 10:33:23 +0000 (13:33 +0300)]
Tonne of indentation + other formatting corrections to rpmdb.c
- no functional changes
Panu Matilainen [Wed, 9 Apr 2008 10:19:26 +0000 (13:19 +0300)]
Remove dead code
Panu Matilainen [Wed, 9 Apr 2008 10:17:01 +0000 (13:17 +0300)]
Report full nevra in rpmdb remove debug message
Panu Matilainen [Wed, 9 Apr 2008 09:47:03 +0000 (12:47 +0300)]
Plug memleak from signature verification now returning malloced msg
Panu Matilainen [Wed, 9 Apr 2008 09:36:38 +0000 (12:36 +0300)]
Simplify addTE()
- eliminate wtf parsing of NEVR string which doesn't actually handle epoch
- use NEVR and NEVRA formatting from headerGetN*() instead of doubling the
work here
Panu Matilainen [Wed, 9 Apr 2008 09:17:36 +0000 (12:17 +0300)]
Make headerGetNEVRA() return "src" as the arch of source packages
Panu Matilainen [Wed, 9 Apr 2008 08:17:43 +0000 (11:17 +0300)]
Initialize pointers, lowercase local variable
Panu Matilainen [Wed, 9 Apr 2008 08:16:24 +0000 (11:16 +0300)]
Rewrite headerGetN*() for simplicity
- eliminate stpcpy() games
- reuse evr formatting from headerGetEVR() instead of doing it over and
over again everywhere
Jindrich Novy [Wed, 9 Apr 2008 08:10:17 +0000 (10:10 +0200)]
Fix buffer overflow in verifyDSASignature()
- caused by assumption that sizeof(size_t) is always 4 (credited to jbj)
Panu Matilainen [Wed, 9 Apr 2008 07:37:34 +0000 (10:37 +0300)]
Replace bunch of wtf evr formatters with headerGetNVR()
Panu Matilainen [Wed, 9 Apr 2008 07:36:29 +0000 (10:36 +0300)]
Add headerGetEVR() convenience function
- return formatted evr string + optionally pointer to name
Panu Matilainen [Wed, 9 Apr 2008 07:32:44 +0000 (10:32 +0300)]
Make headerNEVRA() actually return epoch too
- ...but epoch is an integer type in headers, change proto to match that
Panu Matilainen [Wed, 9 Apr 2008 06:00:44 +0000 (09:00 +0300)]
And one more url->path rename for local variable...
Panu Matilainen [Wed, 9 Apr 2008 05:58:46 +0000 (08:58 +0300)]
Rip the last remaining disk/file URL stuff in processSourceFiles()
Panu Matilainen [Wed, 9 Apr 2008 05:56:27 +0000 (08:56 +0300)]
And more url ripping in build internals... (addFile)
Panu Matilainen [Wed, 9 Apr 2008 05:37:15 +0000 (08:37 +0300)]
URL -> path renames continued... (build internal)
Panu Matilainen [Wed, 9 Apr 2008 05:25:46 +0000 (08:25 +0300)]
Rename FileListRec_s disk + cpio path variables
- on-disk is diskPath, rename previous fileURL to cpioPath to make it
more obvious it's the path in the package cpio
- neither is an URL
Panu Matilainen [Fri, 4 Apr 2008 17:37:48 +0000 (20:37 +0300)]
Yet more url ripping - buildroot is never an url
- this is a bit nasty as it changes public interfaces but rename to make
it explicit they're just plain paths, not urls
Panu Matilainen [Wed, 9 Apr 2008 05:17:48 +0000 (08:17 +0300)]
Remove remote host stuff from build doScript()
Panu Matilainen [Fri, 4 Apr 2008 16:53:40 +0000 (19:53 +0300)]
More url-path ripping in build/
- processBinaryFile() and processMetadataFile() don't need steenking urls
Panu Matilainen [Tue, 8 Apr 2008 14:15:32 +0000 (17:15 +0300)]
Whoops.. gpg-pubkey "release" on import got busted accidentally
Panu Matilainen [Tue, 8 Apr 2008 12:36:46 +0000 (15:36 +0300)]
Simplify rpmtsImportPubkey() a bit
Jindrich Novy [Tue, 8 Apr 2008 11:48:39 +0000 (13:48 +0200)]
Rewrite pgpArmorWrap(), the former one was insecure/hackish
Panu Matilainen [Tue, 8 Apr 2008 11:11:47 +0000 (14:11 +0300)]
Collapse unknown / unsupport RSA sigtypes into one fallthrough
Panu Matilainen [Tue, 8 Apr 2008 11:06:07 +0000 (14:06 +0300)]
verifyRSASignature() buffer & other fixes
- avoid stpcpy() into static sized buffer
- avoid potential sigp NULL-dereferences
Panu Matilainen [Tue, 8 Apr 2008 10:27:59 +0000 (13:27 +0300)]
verifyDSASignature() buffer & other fixes
- avoid stpcpy() into static sized buffer
- avoid potential sigp NULL-dereference
Panu Matilainen [Tue, 8 Apr 2008 10:23:01 +0000 (13:23 +0300)]
verifyDSASignature() buffer & other fixes
- avoid stpcpy() into static sized buffer
- avoid potential sigp NULL-dereference
Panu Matilainen [Tue, 8 Apr 2008 09:35:36 +0000 (12:35 +0300)]
Eliminate four copy-paste versions of nibble(), stick to rpmstring.h
- the one in signature.c wasn't even used...
Panu Matilainen [Mon, 7 Apr 2008 13:13:29 +0000 (16:13 +0300)]
Eep, fix bunch of segfaults and other fun from previous commits
- leftover stpcpy()'s to non-malloced addresses and other "minor"
issues in verify*Signature() stpcpy() elimination ;)
Panu Matilainen [Mon, 7 Apr 2008 12:07:08 +0000 (15:07 +0300)]
Avoid stpcpy() to static sized buffer in verifySizeSignature()
Panu Matilainen [Mon, 7 Apr 2008 12:01:49 +0000 (15:01 +0300)]
Avoid stpcpy() to static sized buffer in verifyMD5Signature()
Panu Matilainen [Mon, 7 Apr 2008 11:55:36 +0000 (14:55 +0300)]
Avoid stpcpy() to static sized buffer in verifySHA1Signature()
Panu Matilainen [Mon, 7 Apr 2008 11:04:00 +0000 (14:04 +0300)]
Eliminate several copy-paste hex converters, use pgpHexStr() instead
Panu Matilainen [Mon, 7 Apr 2008 10:51:20 +0000 (13:51 +0300)]
Uh, return allocated pointer, not the end of string...
Panu Matilainen [Mon, 7 Apr 2008 10:37:56 +0000 (13:37 +0300)]
Remove static print buffer from pgpMpiStr(), return malloc'd data instead
- convert callers
- only rpmpgp.c internal needs, move it there and make static?
Panu Matilainen [Mon, 7 Apr 2008 10:29:48 +0000 (13:29 +0300)]
Merge pgpHexCvt() into pgpHexPrt(), removing it from public headers
Panu Matilainen [Mon, 7 Apr 2008 10:26:46 +0000 (13:26 +0300)]
Minimally convert all pgpHexCvt() users to use pgpHexStr() instead
Panu Matilainen [Mon, 7 Apr 2008 10:02:18 +0000 (13:02 +0300)]
Remove static print buffer from pgpHexStr, return malloc'ed memory instead
- inlined to get xmalloc() from system.h for consistent malloc fail behavior
- convert callers for new behavior, apart from some debug fprintf()'s
Panu Matilainen [Mon, 7 Apr 2008 09:35:37 +0000 (12:35 +0300)]
argvPrint() doesn't modify argv, make it const
Panu Matilainen [Mon, 7 Apr 2008 09:34:07 +0000 (12:34 +0300)]
Add ARGI_const_t, use where appropriate
Panu Matilainen [Mon, 7 Apr 2008 09:11:16 +0000 (12:11 +0300)]
ARGV_t const pendantry
- const on typedef'ed ARGV_t doesn't mean a thing, add a new,
(exec* compatible) ARGV_const_t and use where appropriate consistently in
argv.h, rpmcli.h etc
- popt's argv const is the wrong way around for exec() family, add explicit
ARGV_t casts on popGetArgs() uses where needed
- compiler silence, aaahh...
Panu Matilainen [Mon, 7 Apr 2008 06:31:21 +0000 (09:31 +0300)]
Remove ARGstr_t typedef, it serves no purpose
- argv strings are just regular strings, the structure is what's special
Panu Matilainen [Mon, 7 Apr 2008 06:29:06 +0000 (09:29 +0300)]
Remove bogus const from ARGV_t
Panu Matilainen [Mon, 7 Apr 2008 05:25:17 +0000 (08:25 +0300)]
Missing rpmstring.h include
Panu Matilainen [Sun, 6 Apr 2008 16:34:59 +0000 (19:34 +0300)]
Eliminate the uuuuugly and unsafe static tagname buffer
- generate the "human friendly" strings at build time, store into the tag
table structure
- also avoids copying of strings and lowercasing strings at runtime
Panu Matilainen [Sun, 6 Apr 2008 16:18:44 +0000 (19:18 +0300)]
Move the tagtbl.c generator monster awk out of Makefile.am
- stick it into a separate script, reformat so it's possible to actually
read it...
Panu Matilainen [Fri, 4 Apr 2008 14:00:38 +0000 (17:00 +0300)]
Oops, no newline at end of filename please...
Panu Matilainen [Fri, 4 Apr 2008 13:26:55 +0000 (16:26 +0300)]
%{_builddir} is not an url, ever
Panu Matilainen [Fri, 4 Apr 2008 13:23:56 +0000 (16:23 +0300)]
We only ever deal with local, regular files in doUntar()
Panu Matilainen [Fri, 4 Apr 2008 13:21:58 +0000 (16:21 +0300)]
We only ever deal with local, regular files in doPatch()
Panu Matilainen [Fri, 4 Apr 2008 13:08:37 +0000 (16:08 +0300)]
Rip the url-craziness in rpmMkdirPath()
Panu Matilainen [Fri, 4 Apr 2008 13:03:59 +0000 (16:03 +0300)]
Get rid of the horrid zcmd hackery in build
- rpmExpand() with %uncompress macro etc simplifies things a lot
Panu Matilainen [Fri, 4 Apr 2008 12:52:52 +0000 (15:52 +0300)]
Missing decompression arguments for bzip in uncompress macro
Panu Matilainen [Fri, 4 Apr 2008 12:19:28 +0000 (15:19 +0300)]
getcwd(NULL,0) has undefined behavior, use rpmGetCwd() instead
Panu Matilainen [Fri, 4 Apr 2008 11:59:55 +0000 (14:59 +0300)]
Stop pretending spec can be an URL or stdin, neither works
- simplifies buildForTarget() somewhat
- while at it, eliminate remaining unsafe string operations
Jindrich Novy [Fri, 4 Apr 2008 11:48:02 +0000 (13:48 +0200)]
Escape '%' in format
Panu Matilainen [Fri, 4 Apr 2008 11:33:45 +0000 (14:33 +0300)]
Wait for popen() to finish before passing to isSpecFile()
Panu Matilainen [Fri, 4 Apr 2008 10:52:59 +0000 (13:52 +0300)]
Move spec extraction from tarball out of buildForTarget(), simplify
- use basename() and dirname() instead of manual pointer manipulation games
- static buffer only used for reading tar output, sizes checked
Panu Matilainen [Fri, 4 Apr 2008 09:12:29 +0000 (12:12 +0300)]
Oops, add --wildcards back to spec extraction
- shouldn't really have to hardcode such stuff but...
Jindrich Novy [Fri, 4 Apr 2008 11:17:53 +0000 (13:17 +0200)]
Simplify formatting
Jindrich Novy [Fri, 4 Apr 2008 11:01:50 +0000 (13:01 +0200)]
Simplify package comparison and format creation.
- also avoid static buffer usage
Panu Matilainen [Fri, 4 Apr 2008 09:06:06 +0000 (12:06 +0300)]
Assume failure in buildForTarget() to remove redundancy
Panu Matilainen [Fri, 4 Apr 2008 08:54:21 +0000 (11:54 +0300)]
Refactor spec extraction on tar build to address several issues
- simplify the code
- use %{__tar} instead of tar as the extraction command
- avoid sprintf to "big enough" buffer
- basic spec sanity check while extracting, fixes rhbz#281391
Panu Matilainen [Fri, 4 Apr 2008 08:53:07 +0000 (11:53 +0300)]
Spec of size zero cannot be a valid spec file
Panu Matilainen [Fri, 4 Apr 2008 06:55:30 +0000 (09:55 +0300)]
Remove copy-paste currentDirectory() from sqlite, use rpmGetCwd()
Panu Matilainen [Fri, 4 Apr 2008 06:53:17 +0000 (09:53 +0300)]
Rename currentDirectory() to rpmGetCwd() and move to librpmio
Panu Matilainen [Thu, 3 Apr 2008 17:48:17 +0000 (20:48 +0300)]
We don't support url's in db paths, rip.
- file:// vs / is silly and anything else wont work, so pretending
url support just plain bogus
Panu Matilainen [Thu, 3 Apr 2008 16:44:49 +0000 (19:44 +0300)]
Replace silly static rpmioFileExists() in rpmdb.c with plain old access()
- accepting url's there is not only useless but just plain bogus
Panu Matilainen [Thu, 3 Apr 2008 16:28:37 +0000 (19:28 +0300)]
Remove ancient db1->db3 conversion cruft
Jindrich Novy [Thu, 3 Apr 2008 11:47:05 +0000 (13:47 +0200)]
Don't use static buffers in parseForRegexLang()
- use dynamic allocation instead to avoid overflows
- also update addFile(), which is actually the only function
that calls parseForRegexLang()
Panu Matilainen [Thu, 3 Apr 2008 10:17:57 +0000 (13:17 +0300)]
Lose the remaining static buffer from rpmVerifySignatures()
- Simple and stupid: catenate previous buffer + latest message
with rasprintf() over and over again. Not the most efficient way but hardly
matters here...
Panu Matilainen [Thu, 3 Apr 2008 09:53:11 +0000 (12:53 +0300)]
Typo...
Panu Matilainen [Thu, 3 Apr 2008 09:46:33 +0000 (12:46 +0300)]
Refactor signature check results formatting
- Document what it's supposed to do, should be more obvious from the
code too now.
- Remove one level of stpcpy's to static buffer.
- Simplify the theoretical case where many missing/untrusted keys are
found (this shouldn't happen in practise ever) by only showing last of them
Panu Matilainen [Thu, 3 Apr 2008 06:59:46 +0000 (09:59 +0300)]
Enable stack protector by default if gcc supports it
Panu Matilainen [Thu, 3 Apr 2008 05:37:20 +0000 (08:37 +0300)]
Remove unnecessary check for "keyid:" in result
- nothing in signature.c emits such string
Panu Matilainen [Thu, 3 Apr 2008 05:07:00 +0000 (08:07 +0300)]
Start eliminating static buffers from header/signature checks
- Push msg buffer allocations down to the lowlevel rpmVerifySignature() and
internal verify*Signature functions, nothing above them knows how much
memory they need for messages. Use rasprintf() where obvious,
stupid malloc(bigenuf) for now otherwise.
- Changes public API but can't be helped, printing to an unchecked buffer(s)
of unknown size from one of the more security sensitive pieces is just
.. not ok
- Minimally convert callers to the new allocation scheme
Jindrich Novy [Wed, 2 Apr 2008 16:14:47 +0000 (18:14 +0200)]
Simplify message key creation in i18nTag()
Panu Matilainen [Thu, 3 Apr 2008 05:04:22 +0000 (08:04 +0300)]
Shut up a few silly compiler warnings that have crept in
- missing cast in python header subscript
- (false) warning about uninitialized variable, unused variable in rpmfc
- missing include in rpmio
Panu Matilainen [Wed, 2 Apr 2008 13:52:44 +0000 (16:52 +0300)]
Further streamline / simplify rpmVerifySignatures()
- Use more meaningful variable names than res2 and res3
- In verbose mode we just dump whatever rpmVerifySignature() returns,
error or not. Otherwise the output depends on the result. Make this
obvious in the code.
Panu Matilainen [Wed, 2 Apr 2008 13:30:21 +0000 (16:30 +0300)]
Make signature verification result not depend on rpm verbosity level, ick
Panu Matilainen [Wed, 2 Apr 2008 12:42:21 +0000 (15:42 +0300)]
Further redundancy / copy-paste coding elimination