Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
tim sco5-gcc FAILS (removing ${x86_elf_asm} should fix this) tim I don't think anyone tries to put gnu ld on tim SCO OpenServer 5. In any case, as far as I see we have two choices for 0.9.7: 1. insert the proposed perl filter (it works as stated, and I feel quite confident with it, since Perl works the same everywhere). ... and modify corresponding lines in ./Configure, right? I vote for putting the perl filter in the solaris section. Solaris section? Why are we discussing most *confusing* way to solve it? Unified rule for all ELF platforms (yes, including Solaris) is possible and we should strive for that, not try to overload *kluges*. It won't do anything for SCO OpenServer but it will make UnixWare/OpenUNIX work and probably any other SVR4. (like ReliantUNIX) 2. write a blurb in PROBLEMS stating that for assembler modules to work properly on Intel-based Unixen, it is *required* that GNU as be installed. Which is not 100% true [as there is a kludge implemented for Solaris and it covers just for vendor ass(embler:-)] and most important is somewhat misleading. Because it was established that it was vendor cpp which ruins the build by injecting offending spaces and this would ruin it even for GNU assembler! Blurb should probably be phrased as assembler modules are not *guaranteed* to work on Intel-based none-GNU Unix platforms or something like that (as none-GNU Unix is really a nonsense phrase when you think about it:-). There are two more options. 3. Remove ${x86_elf_asm} from sco5-gcc line. 4. Replace ${x86_elf_asm} in sco5-gcc rule with ${x86_sol_asm} and that would be *it*. No, no extra filter should be required as it will be GNU cpp which will preprocess the .cpp file and no offending spaces will be injected. Yes, I'm contradicting to myself here, but this is not my vote, #3 is. For 0.9.7 that is. A. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
In message [EMAIL PROTECTED] on Fri, 27 Dec 2002 13:30:23 +0100, Andy Polyakov [EMAIL PROTECTED] said: appro There are two more options. appro appro 3. Remove ${x86_elf_asm} from sco5-gcc line. appro appro 4. Replace ${x86_elf_asm} in sco5-gcc rule with ${x86_sol_asm} and that appro would be *it*. No, no extra filter should be required as it will be GNU appro cpp which will preprocess the .cpp file and no offending spaces will be appro injected. Yes, I'm contradicting to myself here, but this is not my appro vote, #3 is. For 0.9.7 that is. Honestly, I personally don't care which is chosen. I think that if it's possible to have it work (with the definition of work I gave before) with assembler code, good. Otherwise, take it out for now, and let's work on that for 0.9.7a or later. -- Richard Levitte \ Spannvägen 38, II \ [EMAIL PROTECTED] Redakteur@Stacken \ S-168 35 BROMMA \ T: +46-8-26 52 47 \ SWEDEN \ or +46-708-26 53 44 Procurator Odiosus Ex Infernis-- [EMAIL PROTECTED] Member of the OpenSSL development team: http://www.openssl.org/ Unsolicited commercial email is subject to an archival fee of $400. See http://www.stacken.kth.se/~levitte/mail/ for more info. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
On Fri, 27 Dec 2002, Andy Polyakov wrote: tim sco5-gcc FAILS (removing ${x86_elf_asm} should fix this) tim I don't think anyone tries to put gnu ld on tim SCO OpenServer 5. In any case, as far as I see we have two choices for 0.9.7: 1. insert the proposed perl filter (it works as stated, and I feel quite confident with it, since Perl works the same everywhere). ... and modify corresponding lines in ./Configure, right? Yes for the UnixWare and OpenUNIX targets. Would probably work for ReliantUNIX but I can't test that. I vote for putting the perl filter in the solaris section. Solaris section? Why are we discussing most *confusing* way to solve it? Unified rule for all ELF platforms (yes, including Solaris) is possible and we should strive for that, not try to overload *kluges*. It won't do anything for SCO OpenServer but it will make UnixWare/OpenUNIX work and probably any other SVR4. (like ReliantUNIX) 2. write a blurb in PROBLEMS stating that for assembler modules to work properly on Intel-based Unixen, it is *required* that GNU as be installed. Which is not 100% true [as there is a kludge implemented for Solaris and it covers just for vendor ass(embler:-)] and most important is somewhat misleading. Because it was established that it was vendor cpp which ruins the build by injecting offending spaces and this would ruin it True. even for GNU assembler! Blurb should probably be phrased as assembler modules are not *guaranteed* to work on Intel-based none-GNU Unix platforms or something like that (as none-GNU Unix is really a nonsense phrase when you think about it:-). There are two more options. 3. Remove ${x86_elf_asm} from sco5-gcc line. This should be done even without any other changes. You'll notice that ${x86_elf_asm} was added to the sco5-gcc line for 0.9.7 and it never worked. 4. Replace ${x86_elf_asm} in sco5-gcc rule with ${x86_sol_asm} and that would be *it*. No, no extra filter should be required as it will be GNU cpp which will preprocess the .cpp file and no offending spaces will be injected. Yes, I'm contradicting to myself here, but this is not my vote, #3 is. For 0.9.7 that is. This wouldn't help the sco5-gcc rule. The vendor as requires -b elf or it will produce COFF objects. -- Tim RiceMultitalents(707) 887-1469 [EMAIL PROTECTED] __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
3. Remove ${x86_elf_asm} from sco5-gcc line. This should be done even without any other changes. You'll notice that ${x86_elf_asm} was added to the sco5-gcc line for 0.9.7 and it never worked. ^^^ I see... 4. Replace ${x86_elf_asm} in sco5-gcc rule with ${x86_sol_asm} and that would be *it*. This wouldn't help the sco5-gcc rule. The vendor as requires -b elf or it will produce COFF objects. I missed this -b elf again:-) In other words we just go for #3 for 0.9.7 and postpone the assembler support [unified ELF rules] for 0.9.7a/8. A. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
On Fri, 27 Dec 2002, Andy Polyakov wrote: 3. Remove ${x86_elf_asm} from sco5-gcc line. This should be done even without any other changes. You'll notice that ${x86_elf_asm} was added to the sco5-gcc line for 0.9.7 and it never worked. ^^^ I see... 4. Replace ${x86_elf_asm} in sco5-gcc rule with ${x86_sol_asm} and that would be *it*. This wouldn't help the sco5-gcc rule. The vendor as requires -b elf or it will produce COFF objects. I missed this -b elf again:-) In other words we just go for #3 for 0.9.7 and postpone the assembler support [unified ELF rules] for 0.9.7a/8. A. If you don't want to put the perl filter in the solaris section so UnixWare can have ASM, then yes option #3. I'll just use a simple local patch here until 0.9.7a. -- Tim RiceMultitalents(707) 887-1469 [EMAIL PROTECTED] __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
3. Remove ${x86_elf_asm} from sco5-gcc line. Feels like a wrap-up point for me... I've verified solaris[64]-sparcv9-[g]cc, solaris-x86-gcc, irix[64]-mips3-cc, linux-alpha-[gc]cc, linux-x86_64 and linux-ia64(*) targets. Well, not to mention linux-pentium and FreeBSD-elf... As for DES assembler PIC-ification. There were issues with all Win32 assemblers [masm, nasm, cygwin] which were resolved in cooperation with Nathan Kidd [of Hummingbird] *before* beta6 was out so that the transition is believed to be complete. STATUS mentions Windows 2000 (Cygwin) ERROR (PR: 404). Should I try to reproduce it? Well, to me it actually sounds more like environmental problem (there're these hard-to-figure-out-for-a-unix-person binmode and textmode mount options in cygwin) and I'd simply resolve it as not really an OpenSSL problem. STATUS mentions SCO OpenServer, Unixware? ERROR. As was mentioned earlier, sco5-gcc is actually the only one that fails which is now resolved by stripping the assembler support. BN_kronecker from RT#410 is my only worry... I failed to reproduce the problem on irix-mips3-cc with no-asm which is architecturally similar to hpux-parisc2-cc being discussed in this ticket. At least former [irix-mips3-cc that is] turned to exhibit originally reported failure in BN_sqr and posted patch was developed and checked on it. Cheers. A. (*) Well, dsatest fails on *me*, but I have reasons to believe it's caused by a[nother] bug in gcc which comes with this HP NUE emulator thing I'm using. That's because a late pre-beta6 snapshot was reported to pass the test suite on a real IA64 Suse distro. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
In message [EMAIL PROTECTED] on Wed, 25 Dec 2002 21:03:37 -0800 (PST), Tim Rice [EMAIL PROTECTED] said: tim I'll try and work on it between getting a server ready to install on monday. tim I've attached a couple of sed man pages. (in case you figure it out first) Hmm, they both refer to ed for regular expressions... -- Richard Levitte \ Spannvägen 38, II \ [EMAIL PROTECTED] Redakteur@Stacken \ S-168 35 BROMMA \ T: +46-8-26 52 47 \ SWEDEN \ or +46-708-26 53 44 Procurator Odiosus Ex Infernis-- [EMAIL PROTECTED] Member of the OpenSSL development team: http://www.openssl.org/ Unsolicited commercial email is subject to an archival fee of $400. See http://www.stacken.kth.se/~levitte/mail/ for more info. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
On Thu, 26 Dec 2002, Richard Levitte - VMS Whacker wrote: In message [EMAIL PROTECTED] on Wed, 25 Dec 2002 21:03:37 -0800 (PST), Tim Rice [EMAIL PROTECTED] said: tim I'll try and work on it between getting a server ready to install on monday. tim I've attached a couple of sed man pages. (in case you figure it out first) Hmm, they both refer to ed for regular expressions... Those last minute e-mails before crawling off to bed will get you every time. I've attached the ed man pages. Perhaps we are going down the wrong path. Instead of wasting our time trying to figure out each platform's sed, maybe we should be using perl. Perl should work the same on all platforms. Unfortunately neither of these work perl -pi -e 's/ \?\([\.,@]\) */\1/g;' -e 's/ *:/:/g;' -e 's/#.*//' /tmp/testfile perl -pi -e 's/ +\([\.,:@]\) +/\1/g;' -e 's/#.*//' /tmp/testfile for lines like . type md5_block_asm_host_order , @ function -- Tim RiceMultitalents(707) 887-1469 [EMAIL PROTECTED] ed(C) 19 June 1992 ed(C) Name ed, red - invoke the text editor Syntax ed [ - ] [ -p string ] [ file ] red [ - ] [ -p string ] [ file ] Description ed is the standard text editor. If the file argument is given, ed simu- lates an e command (see below) on the named file; that is to say, the file is read into ed's buffer so that it can be edited. ed operates on a copy of the file it is editing; changes made to the copy have no effect on the file until a w (write) command is given. The copy of the text being edited resides in a temporary file called the buffer. There is only one buffer. red is a restricted version of ed(C). It will only allow editing of files in the current directory. It prohibits executing sh(C) commands via the ! command. red displays an error message on any attempt to bypass these restrictions. In general, red does not allow commands like !date or !sh. Furthermore, red will not allow pathnames in its command line. For exam- ple, the command: red /etc/passwd when the current directory is not /etc causes an error. Options The options to ed are: - Suppresses the printing of character counts by the e, r, and w commands, of diagnostics from e and q commands, and the ``!'' prompt after a ! shell command. -p Allows the user to specify a prompt string. ed supports formatting capability. After including a format specifica- tion as the first line of file and invoking ed with your terminal in stty-tabs or sttytab3 mode (see stty(C)), the specified tab stops will automatically be used when scanning file. For example, if the first line of a file contained: :t5,10,15 s72: tab stops would be set at columns 5, 10, and 15, and a maximum line length of 72 would be imposed. Note: While inputting text, tab characters are expanded to every eighth column as the default. Commands to ed have a simple and regular structure: zero, one, or two addresses followed by a single-character command, possibly followed by parameters to that command. These addresses specify one or more lines in the buffer. Every command that requires addresses has default addresses, so that the addresses can very often be omitted. In general, only one command may appear on a line. Certain commands allow the input of text. This text is placed in the appropriate place in the buffer. While ed is accepting text, it is said to be in input mode. In this mode, no commands are recognized; all input is merely collected. Input mode is left by entering a period (.) alone at the beginning of a line. ed supports a limited form of regular expression notation; regular expressions are used in addresses to specify lines and in some commands (for example, s) to specify portions of a line that are to be substi- tuted. A regular expression specifies a set of character strings. A member of this set of strings is said to be matched by the regular expression. The regular expressions allowed by ed are constructed as follows: The following one-character regular expressions match a single character: 1.1 An ordinary character (not one of those discussed in 1.2 below) is a one-character regular expression that matches itself. 1.2 A backslash (\) followed by any special character is a one-character regular expression that matches the special character itself. The special characters are: a. . * [ and \ (dot, star, left square bracket, and backslash, respectively), which are otherwise special, except when they appear within square
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
tim I've attached a couple of sed man pages. (in case you figure it out first) Hmm, they both refer to ed for regular expressions... They used to have manual pages on-line, but apparently it's moved... Aha! See http://www.caldera.com/support/docs/. No need to bore the whole list with attachements only a couple of persons will read:-) Perhaps we are going down the wrong path. Perhaps... maybe we should be using perl. If it ought to be perl, then I'd rather get rid of $(CPP) altogether. In which case I'd implement elf-pic perlasm option and simply asm/dx586-elf.o:asm/dx586-elf.s $(CC) $(CFLAGS) -o $@ $ asm/dx586-elf.s:asm/des-586.pl (cd asm; $(PERL) des-586.pl `echo elf or elf-pic` dx586-elf.s) and make sure perl emits universal code (i.e. no comments:-). Note that the first rule automatically covers for -b elf of yours:-) Cheers. A. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
In message [EMAIL PROTECTED] on Thu, 26 Dec 2002 09:36:09 -0800 (PST), Tim Rice [EMAIL PROTECTED] said: tim Perhaps we are going down the wrong path. Instead of wasting our time tim trying to figure out each platform's sed, maybe we should be using perl. tim Perl should work the same on all platforms. You're right, perl is more reliable. tim Unfortunately neither of these work timperl -pi -e 's/ \?\([\.,@]\) */\1/g;' -e 's/ *:/:/g;' -e 's/#.*//' /tmp/testfile timperl -pi -e 's/ +\([\.,:@]\) +/\1/g;' -e 's/#.*//' /tmp/testfile tim for lines like tim . type md5_block_asm_host_order , @ function The syntax for perl regexps is a little bit different. This works: perl -pi -e 's/ ?([\.,@]) +/\1/g; s/ +:/:/g; s/#.*//;' Here are a few test lines (paired with results): . type md5_block_asm_host_order , @ function .type md5_block_asm_host_order,@function foo : foo: foo : # bah foo: -- Richard Levitte \ Spannvägen 38, II \ [EMAIL PROTECTED] Redakteur@Stacken \ S-168 35 BROMMA \ T: +46-8-26 52 47 \ SWEDEN \ or +46-708-26 53 44 Procurator Odiosus Ex Infernis-- [EMAIL PROTECTED] Member of the OpenSSL development team: http://www.openssl.org/ Unsolicited commercial email is subject to an archival fee of $400. See http://www.stacken.kth.se/~levitte/mail/ for more info. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
In message [EMAIL PROTECTED] on Thu, 26 Dec 2002 21:08:00 +0100, Andy Polyakov [EMAIL PROTECTED] said: appro maybe we should be using perl. appro appro If it ought to be perl, then I'd rather get rid of $(CPP) altogether. In appro which case I'd implement elf-pic perlasm option and simply appro appro asm/dx586-elf.o: asm/dx586-elf.s appro $(CC) $(CFLAGS) -o $@ $ appro appro asm/dx586-elf.s: asm/des-586.pl appro (cd asm; $(PERL) des-586.pl `echo elf or elf-pic` dx586-elf.s) appro appro and make sure perl emits universal code (i.e. no comments:-). Note that appro the first rule automatically covers for -b elf of yours:-) How about reworking that for 0.9.8? Or if you dare, for 0.9.7a? -- Richard Levitte \ Spannvägen 38, II \ [EMAIL PROTECTED] Redakteur@Stacken \ S-168 35 BROMMA \ T: +46-8-26 52 47 \ SWEDEN \ or +46-708-26 53 44 Procurator Odiosus Ex Infernis-- [EMAIL PROTECTED] Member of the OpenSSL development team: http://www.openssl.org/ Unsolicited commercial email is subject to an archival fee of $400. See http://www.stacken.kth.se/~levitte/mail/ for more info. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
appro maybe we should be using perl. appro appro If it ought to be perl, then I'd rather get rid of $(CPP) altogether. In appro which case I'd implement elf-pic perlasm option and simply appro appro asm/dx586-elf.o: asm/dx586-elf.s appro $(CC) $(CFLAGS) -o $@ $ appro appro asm/dx586-elf.s: asm/des-586.pl appro (cd asm; $(PERL) des-586.pl `echo elf or elf-pic` dx586-elf.s) appro appro and make sure perl emits universal code (i.e. no comments:-). Note that appro the first rule automatically covers for -b elf of yours:-) How about reworking that for 0.9.8? Or if you dare, for 0.9.7a? ??? I already said that I'm holding this [universal rule for all IA-32 ELF targets] for 0.9.7a... You don't have to dare me... A:-) __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
In message [EMAIL PROTECTED] on Thu, 26 Dec 2002 22:02:49 +0100, Andy Polyakov [EMAIL PROTECTED] said: appro appro and make sure perl emits universal code (i.e. no comments:-). Note that appro appro the first rule automatically covers for -b elf of yours:-) appro appro How about reworking that for 0.9.8? Or if you dare, for 0.9.7a? appro appro ??? I already said that I'm holding this [universal rule for all IA-32 appro ELF targets] for 0.9.7a... You don't have to dare me... A:-) I believe that a swicth from using $(CPP) followed by a rather simple filter to a solution completely in Perl is a rather large step. Therein lies the dare. Remember that we don't do betas on patch levels since a bit... In the mean time, does the current solution work on UnixWare/SCO or doesn't it? If it doesn't, and the simple filter makes it work, I'd say we do it, or we have a showstopper, IMHO. -- Richard Levitte \ Spannvägen 38, II \ [EMAIL PROTECTED] Redakteur@Stacken \ S-168 35 BROMMA \ T: +46-8-26 52 47 \ SWEDEN \ or +46-708-26 53 44 Procurator Odiosus Ex Infernis-- [EMAIL PROTECTED] Member of the OpenSSL development team: http://www.openssl.org/ Unsolicited commercial email is subject to an archival fee of $400. See http://www.stacken.kth.se/~levitte/mail/ for more info. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
appro appro and make sure perl emits universal code (i.e. no comments:-). Note that appro appro the first rule automatically covers for -b elf of yours:-) appro appro How about reworking that for 0.9.8? Or if you dare, for 0.9.7a? appro appro ??? I already said that I'm holding this [universal rule for all IA-32 appro ELF targets] for 0.9.7a... You don't have to dare me... A:-) I believe that a swicth from using $(CPP) followed by a rather simple filter to a solution completely in Perl is a rather large step. Therein lies the dare. Remember that we don't do betas on patch levels since a bit... Yet I'd dare:-) If it was up to me and if Tim is with us [i.e. is ready to swiftly verify a snapshot on explicit request], I'd pull it [unified *586-elf.o rules] even now:-) I have Linux and Solaris/Intel, login.openssl.org is a FreeBSD machine... So shall we? In the mean time, does the current solution work on UnixWare/SCO or doesn't it? Define current solution and define work. As far as I can tell the only target that shouldn't work in 0.9.7-beta6 is sco5-gcc and only if you don't have GNU assembler around. The rest of targets should work. They lack assembler support, but they're operational. What is being discussed here is *proposed* patch adding support for assembler modules to the targets in question. A. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
In message [EMAIL PROTECTED] on Thu, 26 Dec 2002 23:31:38 +0100, Andy Polyakov [EMAIL PROTECTED] said: appro Yet I'd dare:-) If it was up to me and if Tim is with us [i.e. is ready appro to swiftly verify a snapshot on explicit request], I'd pull it [unified appro *586-elf.o rules] even now:-) I have Linux and Solaris/Intel, appro login.openssl.org is a FreeBSD machine... So shall we? I suggest not. Can you say that you cover for all possible variants of Linux? All variants of FreeBSD? What about OpenBSD? What about NetBSD? I know that the *probability* is that it will work fine everywhere. Unfortunately, when I've tried probability, I've been bitten a few times. I'd therefore rather play it safe. Remember that the plan is to release in the next few days. Do you want to take responsability for the support requests because whatever you do doesn't work on platform FOO with compiler BAR and linker BAZ? appro In the mean time, does the current solution work on UnixWare/SCO or appro doesn't it? appro appro Define current solution and define work. current solution: what is provided with OpenSSL 0.9.7 beta7-dev and gets configured by default. works: goes through compilation with no errors or warnings, goes through tests with no errors (except for expired certificates in some of the tests), installs correctly, and performs correctly with applications compiled against it. All this, of course, as far as we're able to test (which often makes the part about correctly working applications tricky, and therefore usually just an option, since we haven't even defined what software we expect to run properly). So if that means that currently there are some configuration targets that do not provide any assembler modules and therefore use C code only, and works properly with the definition above, that's perfectly fine. In my opinion, works comes before performs faster (which is the expectation with the assembler modules :-)). appro As far as I can tell the only target that shouldn't work in appro 0.9.7-beta6 is sco5-gcc and only if you don't have GNU appro assembler around. The rest of targets should work. They lack appro assembler support, but they're operational. What is being appro discussed here is *proposed* patch adding support for assembler appro modules to the targets in question. OK, now I have a better picture. For some reason, I didn't see the beginning of this thread. If sco5-gcc currently doesn't use any assembler and works with such a configuration, I'd also leave it be until 0.9.7a. -- Richard Levitte \ Spannvägen 38, II \ [EMAIL PROTECTED] Redakteur@Stacken \ S-168 35 BROMMA \ T: +46-8-26 52 47 \ SWEDEN \ or +46-708-26 53 44 Procurator Odiosus Ex Infernis-- [EMAIL PROTECTED] Member of the OpenSSL development team: http://www.openssl.org/ Unsolicited commercial email is subject to an archival fee of $400. See http://www.stacken.kth.se/~levitte/mail/ for more info. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
appro Yet I'd dare:-) If it was up to me and if Tim is with us [i.e. is ready appro to swiftly verify a snapshot on explicit request], I'd pull it [unified appro *586-elf.o rules] even now:-) I have Linux and Solaris/Intel, appro login.openssl.org is a FreeBSD machine... So shall we? I suggest not. Oops! Missed smiley after so shall we. I mean I'm not actually suggesting to pull it *now*. But I'm serious about 0.9.7a... appro As far as I can tell the only target that shouldn't work in appro 0.9.7-beta6 is sco5-gcc and only if you don't have GNU appro assembler around. The rest of targets should work. They lack appro assembler support, but they're operational. What is being appro discussed here is *proposed* patch adding support for assembler appro modules to the targets in question. OK, now I have a better picture. For some reason, I didn't see the beginning of this thread. If sco5-gcc currently doesn't use any assembler and works with such a configuration, Well, I meant that sco5-gcc is the only one [among ~16 being discussed] that *uses* assembler modules and *therefore* might not work *unless* gcc is configured to invoke gnu assembler. I don't know if gnu assembler is mandatory for gcc installation on SCO or is it optional like in Solaris. But in either case we can always point at no-asm. A. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
On Thu, 26 Dec 2002, Richard Levitte - VMS Whacker wrote: levitte In message [EMAIL PROTECTED] on Thu, 26 Dec 2002 23:31:38 +0100, Andy Polyakov [EMAIL PROTECTED] said: levitte levitte appro Yet I'd dare:-) If it was up to me and if Tim is with us [i.e. is ready levitte appro to swiftly verify a snapshot on explicit request], I'd pull it [unified levitte appro *586-elf.o rules] even now:-) I have Linux and Solaris/Intel, levitte appro login.openssl.org is a FreeBSD machine... So shall we? levitte levitte I suggest not. Can you say that you cover for all possible variants levitte of Linux? All variants of FreeBSD? What about OpenBSD? What about levitte NetBSD? I'm ready to test. But Richard is right, we're talking about a major change. [snip] levitte appro As far as I can tell the only target that shouldn't work in levitte appro 0.9.7-beta6 is sco5-gcc and only if you don't have GNU levitte appro assembler around. The rest of targets should work. They lack levitte appro assembler support, but they're operational. What is being levitte appro discussed here is *proposed* patch adding support for assembler levitte appro modules to the targets in question. levitte levitte OK, now I have a better picture. For some reason, I didn't see the levitte beginning of this thread. If sco5-gcc currently doesn't use any levitte assembler and works with such a configuration, I'd also leave it be levitte until 0.9.7a. Here is where we stand. Without adding my patch to support ASM on UnixWare and OpenServer The UnixWare cc targets are OK sco3-gccOK sco5-cc and sco5-cc-pentiun OK sco5-gccFAILS (removing ${x86_elf_asm} should fix this) I don't think anyone tries to put gnu ld on SCO OpenServer 5. I'd really like to see the ... sed 's/^#.*//' ... in openssl-0.9.7/crypto/bf/Makefile.ssl openssl-0.9.7/crypto/bn/Makefile.ssl openssl-0.9.7/crypto/cast/Makefile.ssl openssl-0.9.7/crypto/des/Makefile.ssl openssl-0.9.7/crypto/md5/Makefile.ssl openssl-0.9.7/crypto/rc4/Makefile.ssl openssl-0.9.7/crypto/rc5/Makefile.ssl openssl-0.9.7/crypto/ripemd/Makefile.ssl openssl-0.9.7/crypto/sha/Makefile.ssl replaced with ... perl -pi -e 's/ ?([\.,@]) +/\1/g; s/ +:/:/g; s/#.*//;' ... in 0.9.7 Then we could add ${x86_sol_asm} to the UnixWare/OpenUNIX targets giving them asm support. We can leave the major rework to 0.9.7a -- Tim RiceMultitalents(707) 887-1469 [EMAIL PROTECTED] __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
A lot of developers have waited a very long time for 0.9.7, over two years, and now that its in final beta tests, a discussion is active about making major changes to it. Does not make sense. Ken On Thu, 26 Dec 2002, Richard Levitte - VMS Whacker wrote: levitte In message [EMAIL PROTECTED] on Thu, 26 Dec 2002 23:31:38 +0100, Andy Polyakov [EMAIL PROTECTED] said: levitte levitte appro Yet I'd dare:-) If it was up to me and if Tim is with us [i.e. is ready levitte appro to swiftly verify a snapshot on explicit request], I'd pull it [unified levitte appro *586-elf.o rules] even now:-) I have Linux and Solaris/Intel, levitte appro login.openssl.org is a FreeBSD machine... So shall we? levitte levitte I suggest not. Can you say that you cover for all possible variants levitte of Linux? All variants of FreeBSD? What about OpenBSD? What about levitte NetBSD? I'm ready to test. But Richard is right, we're talking about a major change. [snip] levitte appro As far as I can tell the only target that shouldn't work in levitte appro 0.9.7-beta6 is sco5-gcc and only if you don't have GNU levitte appro assembler around. The rest of targets should work. They lack levitte appro assembler support, but they're operational. What is being levitte appro discussed here is *proposed* patch adding support for assembler levitte appro modules to the targets in question. levitte levitte OK, now I have a better picture. For some reason, I didn't see the levitte beginning of this thread. If sco5-gcc currently doesn't use any levitte assembler and works with such a configuration, I'd also leave it be levitte until 0.9.7a. Here is where we stand. Without adding my patch to support ASM on UnixWare and OpenServer The UnixWare cc targets are OK sco3-gcc OK sco5-cc and sco5-cc-pentiun OK sco5-gcc FAILS (removing ${x86_elf_asm} should fix this) I don't think anyone tries to put gnu ld on SCO OpenServer 5. I'd really like to see the ... sed 's/^#.*//' ... in openssl-0.9.7/crypto/bf/Makefile.ssl openssl-0.9.7/crypto/bn/Makefile.ssl openssl-0.9.7/crypto/cast/Makefile.ssl openssl-0.9.7/crypto/des/Makefile.ssl openssl-0.9.7/crypto/md5/Makefile.ssl openssl-0.9.7/crypto/rc4/Makefile.ssl openssl-0.9.7/crypto/rc5/Makefile.ssl openssl-0.9.7/crypto/ripemd/Makefile.ssl openssl-0.9.7/crypto/sha/Makefile.ssl replaced with ... perl -pi -e 's/ ?([\.,@]) +/\1/g; s/ +:/:/g; s/#.*//;' ... in 0.9.7 Then we could add ${x86_sol_asm} to the UnixWare/OpenUNIX targets giving them asm support. We can leave the major rework to 0.9.7a -- Tim Rice Multitalents(707) 887-1469 [EMAIL PROTECTED] __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED] __ Support InterSoft International, Inc. Voice: 888-823-1541, International 281-398-7060 Fax: 888-823-1542, International 281-398-0221 [EMAIL PROTECTED] http://www.securenetterm.com __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
Let me make sure I understand. One SCO Unix, GCC without GNU ASM *might* require -no-asm. That's the risk. The benefit is much cleaner make. I say go for it. /r$ __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
In message 3E0B49CA.4377.9884857@localhost on Thu, 26 Dec 2002 18:26:18 -0600, Kenneth R. Robinette [EMAIL PROTECTED] said: support A lot of developers have waited a very long time for 0.9.7, over two years, and now support that its in final beta tests, a discussion is active about making major changes to it. support Does not make sense. I entirely agree with you. However, it seems like that's discussions for the future, mostly. However, considering we still have unfixed errors on some supported platforms (Windows), it looks like the release won't happen within the specified time range. I don't expect it to be delayed with more than a week... -- Richard Levitte \ Spannvägen 38, II \ [EMAIL PROTECTED] Redakteur@Stacken \ S-168 35 BROMMA \ T: +46-8-26 52 47 \ SWEDEN \ or +46-708-26 53 44 Procurator Odiosus Ex Infernis-- [EMAIL PROTECTED] Member of the OpenSSL development team: http://www.openssl.org/ Unsolicited commercial email is subject to an archival fee of $400. See http://www.stacken.kth.se/~levitte/mail/ for more info. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
In message [EMAIL PROTECTED] on Thu, 26 Dec 2002 16:17:28 -0800 (PST), Tim Rice [EMAIL PROTECTED] said: tim sco5-gcc FAILS (removing ${x86_elf_asm} should fix this) timI don't think anyone tries to put gnu ld on timSCO OpenServer 5. Uhmm, we're talking gnu as, not gnu ld. Does your statement still stand? Also, are there other SCO users on this list with different experiences, or is it the same for everyone? In any case, as far as I see we have two choices for 0.9.7: 1. insert the proposed perl filter (it works as stated, and I feel quite confident with it, since Perl works the same everywhere). 2. write a blurb in PROBLEMS stating that for assembler modules to work properly on Intel-based Unixen, it is *required* that GNU as be installed. I can go for either of them. -- Richard Levitte \ Spannvägen 38, II \ [EMAIL PROTECTED] Redakteur@Stacken \ S-168 35 BROMMA \ T: +46-8-26 52 47 \ SWEDEN \ or +46-708-26 53 44 Procurator Odiosus Ex Infernis-- [EMAIL PROTECTED] Member of the OpenSSL development team: http://www.openssl.org/ Unsolicited commercial email is subject to an archival fee of $400. See http://www.stacken.kth.se/~levitte/mail/ for more info. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
Whats wrong with Windows? Ken In message 3E0B49CA.4377.9884857@localhost on Thu, 26 Dec 2002 18:26:18 -0600, Kenneth R. Robinette [EMAIL PROTECTED] said: support A lot of developers have waited a very long time for 0.9.7, over two years, and now support that its in final beta tests, a discussion is active about making major changes to it. support Does not make sense. I entirely agree with you. However, it seems like that's discussions for the future, mostly. However, considering we still have unfixed errors on some supported platforms (Windows), it looks like the release won't happen within the specified time range. I don't expect it to be delayed with more than a week... -- Richard Levitte \ Spannvägen 38, II \ [EMAIL PROTECTED] Redakteur@Stacken \ S-168 35 BROMMA \ T: +46-8-26 52 47 \ SWEDEN \ or +46-708-26 53 44 Procurator Odiosus Ex Infernis-- [EMAIL PROTECTED] Member of the OpenSSL development team: http://www.openssl.org/ Unsolicited commercial email is subject to an archival fee of $400. See http://www.stacken.kth.se/~levitte/mail/ for more info. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED] __ Support InterSoft International, Inc. Voice: 888-823-1541, International 281-398-7060 Fax: 888-823-1542, International 281-398-0221 [EMAIL PROTECTED] http://www.securenetterm.com __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
On Fri, 27 Dec 2002, Richard Levitte - VMS Whacker wrote: In message [EMAIL PROTECTED] on Thu, 26 Dec 2002 16:17:28 -0800 (PST), Tim Rice [EMAIL PROTECTED] said: tim sco5-gcc FAILS (removing ${x86_elf_asm} should fix this) tim I don't think anyone tries to put gnu ld on tim SCO OpenServer 5. Uhmm, we're talking gnu as, not gnu ld. Does your statement still stand? Duh, yes my statement still stands. Sorry, most of my focus is on getting a server ready to install Monday. Also, are there other SCO users on this list with different experiences, or is it the same for everyone? I wonder if there is even enyone else on the list testing on SCO OpenServer. None of the 0.9.7 sco builds have worked until about a week ago and I've not seen any report from any other SCO user. Any SCO users out there? In any case, as far as I see we have two choices for 0.9.7: 1. insert the proposed perl filter (it works as stated, and I feel quite confident with it, since Perl works the same everywhere). I vote for putting the perl filter in the solaris section. It won't do anything for SCO OpenServer but it will make UnixWare/OpenUNIX work and probably any other SVR4. (like ReliantUNIX) 2. write a blurb in PROBLEMS stating that for assembler modules to work properly on Intel-based Unixen, it is *required* that GNU as be installed. I can go for either of them. -- Tim RiceMultitalents(707) 887-1469 [EMAIL PROTECTED] __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
In message 3E0B85E7.26469.A731912@localhost on Thu, 26 Dec 2002 22:42:47 -0600, Kenneth R. Robinette [EMAIL PROTECTED] said: support Whats wrong with Windows? There's a report that it doesn't work with MIT Kerberos (which we claim to support). I'm thinking we shouldn't consider that a showstopper. -- Richard Levitte \ Spannvägen 38, II \ [EMAIL PROTECTED] Redakteur@Stacken \ S-168 35 BROMMA \ T: +46-8-26 52 47 \ SWEDEN \ or +46-708-26 53 44 Procurator Odiosus Ex Infernis-- [EMAIL PROTECTED] Member of the OpenSSL development team: http://www.openssl.org/ Unsolicited commercial email is subject to an archival fee of $400. See http://www.stacken.kth.se/~levitte/mail/ for more info. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
In message [EMAIL PROTECTED] on Tue, 24 Dec 2002 16:21:32 -0800 (PST), Tim Rice [EMAIL PROTECTED] said: tim sed -e 's/ \?\([\.,@]\) */\1/g' -e 's/ *:/:/g' -e 's/#.*//' tim tim It doesn't work with the sed on UnixWare or sed on SCO OpenServer. tim The 's/#.*//' rule works. tim It seems to do the right thing with GNU sed. Do you have the possibility to either figure out a sed expression that works with the SCO and UnixWare seds, or send me the man pages for sed on each of those systems? As far as I understand, we need the filtering I picture above. Do you understand what it does? -- Richard Levitte \ Spannvägen 38, II \ [EMAIL PROTECTED] Redakteur@Stacken \ S-168 35 BROMMA \ T: +46-8-26 52 47 \ SWEDEN \ or +46-708-26 53 44 Procurator Odiosus Ex Infernis-- [EMAIL PROTECTED] Member of the OpenSSL development team: http://www.openssl.org/ Unsolicited commercial email is subject to an archival fee of $400. See http://www.stacken.kth.se/~levitte/mail/ for more info. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
On Wed, 25 Dec 2002, Richard Levitte - VMS Whacker wrote: In message [EMAIL PROTECTED] on Tue, 24 Dec 2002 16:21:32 -0800 (PST), Tim Rice [EMAIL PROTECTED] said: tim sed -e 's/ \?\([\.,@]\) */\1/g' -e 's/ *:/:/g' -e 's/#.*//' tim tim It doesn't work with the sed on UnixWare or sed on SCO OpenServer. tim The 's/#.*//' rule works. tim It seems to do the right thing with GNU sed. Do you have the possibility to either figure out a sed expression that works with the SCO and UnixWare seds, or send me the man pages for sed on each of those systems? I'll try and work on it between getting a server ready to install on monday. I've attached a couple of sed man pages. (in case you figure it out first) As far as I understand, we need the filtering I picture above. Do you understand what it does? I'll need some words explaining what the first section should do. -- Tim RiceMultitalents(707) 887-1469 [EMAIL PROTECTED] __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
Rats, forgot to attach the man pages. On Wed, 25 Dec 2002, Tim Rice wrote: On Wed, 25 Dec 2002, Richard Levitte - VMS Whacker wrote: In message [EMAIL PROTECTED] on Tue, 24 Dec 2002 16:21:32 -0800 (PST), Tim Rice [EMAIL PROTECTED] said: tim sed -e 's/ \?\([\.,@]\) */\1/g' -e 's/ *:/:/g' -e 's/#.*//' tim tim It doesn't work with the sed on UnixWare or sed on SCO OpenServer. tim The 's/#.*//' rule works. tim It seems to do the right thing with GNU sed. Do you have the possibility to either figure out a sed expression that works with the SCO and UnixWare seds, or send me the man pages for sed on each of those systems? I'll try and work on it between getting a server ready to install on monday. I've attached a couple of sed man pages. (in case you figure it out first) As far as I understand, we need the filtering I picture above. Do you understand what it does? I'll need some words explaining what the first section should do. -- Tim RiceMultitalents(707) 887-1469 [EMAIL PROTECTED] sed(C)06 January 1993 sed(C) Name sed - invoke the stream editor Syntax sed [ -n ] [ -e script ] [ -f sfile ] [ files ] Description The sed command copies the named files (standard input default) to the standard output, edited according to a script of commands. The -e option causes the script to be read literally from the next argument, which is usually quoted to protect it from the shell. The -f option causes the script to be taken from file sfile; these options accumulate. If there is just one -e option and no -f options, the flag -e may be omitted. The -n option suppresses the default output. A script consists of editing com- mands, one per line, of the following form: [ address [ , address ] ] function [ arguments ] In normal operation, sed cyclically copies a line of input into a pattern space (unless there is something left after a D command), applies in sequence all commands whose addresses select that pattern space, and at the end of the script copies the pattern space to the standard output (except under -n) and deletes the pattern space. A semicolon (;) can be used as a command delimiter. Some of the commands use a hold space to save all or part of the pattern space for subsequent retrieval. An address is either a decimal number that counts input lines cumula- tively across files, a ``$'' that addresses the last line of input, or a context address, that is, a /regular expression/ in the style of ed(C) modified as follows: + In a context address, the construction \?regular expression?, where ``?'' is any character, is identical to /regular expression/. Note that in the context address \xabc\xdefx, the second x stands for itself, so that the standard expression is abcxdef. + The escape sequence \n matches a newline embedded in the pattern space. + A dot (.) matches any character except the terminal newline of the pattern space. + A command line with no addresses selects every pattern space. + A command line with one address selects each pattern space that matches the address. + A command line with two addresses separated by a comma selects the inclusive range from the first pattern space that matches the first address through the next pattern space that matches the second. (If the second address is a number less than or equal to the line number first selected, only one line is selected.) Thereafter, the process is repeated, looking again for the first address. Editing commands can be applied only to nonselected pattern spaces by use of the negation function ``!'' (below). In the following list of functions, the maximum number of permissible addresses for each function is indicated in parentheses. The text argument consists of one or more lines, all but the last of which end with backslashes to hide the newlines. Backslashes in text are treated like backslashes in the replacement string of an s command, and may be used to protect initial blanks and tabs against the stripping that is done on every script line. The rfile or wfile argument must terminate the command line and must be preceded by one blank. Each wfile is created before processing begins. There can be at most 10 distinct wfile argu- ments. (1) a\ text Appends text, placing it on the output before reading the next input line. (2) b label Branches to the : command bearing the label. If label is empty, branches to the end of the script. (2) c\ text Changes text by deleting the
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
On Mon, 23 Dec 2002, Andy Polyakov wrote: My solution was to add support for assembly modules. We probably have to postpone this patch to 0.9.7a. If you only could reply more swiftly so that the changes could be exposed in beta... Sorry for the delay. I've been doing some paying jobs so I can pay bills. :-) [snip] Is it $(CPP) in question which adds them? Yes, $(CPP) adds the spaces. I can imagine / :/:/, but where do /\. /./ and /@ /@/ some from? Can you pinpoint the lines? I mean just give the offending lines' number... Here are the erors from UnixWare 7.1.1 ... (cd asm; /usr/bin/perl md5-586.pl cpp mx86unix.cpp) cc -E -DELF asm/mx86unix.cpp asm/mx86svr4.s as -o asm/mx86svr4.o asm/mx86svr4.s rm -f asm/mx86svr4.s UX:as: ERROR: asm/mx86svr4.s:33:syntax error at name: md5_block_asm_host_order UX:as: ERROR: asm/mx86svr4.s:33:invalid @ operator UX:as: ERROR: asm/mx86svr4.s:707:syntax error at name: md5_block_asm_host_order ... Here is the 2 chunks of code ... .align 16 globl md5_block_asm_host_order . type md5_block_asm_host_order , @ function md5_block_asm_host_order: pushl %esi ... ret L_md5_block_asm_host_order_end: . size md5_block_asm_host_order , . L_md5_block_asm_host_order_end - md 5_block_asm_host_order ident desasm.pl #ident acomp: (CCS) 3.2 09/28/99 (santamariasbl5.1.rs1ubl7) ... Now I remove the first space .type md5_block_asm_host_order , @ function and do as -o asm/mx86svr4.o asm/mx86svr4.s, I see tim@uw711 74% as -o asm/mx86svr4.o asm/mx86svr4.s UX:as: ERROR: asm/mx86svr4.s:33:invalid @ operator UX:as: ERROR: asm/mx86svr4.s:33:syntax error at @GOTOFF UX:as: ERROR: asm/mx86svr4.s:707:syntax error at name: md5_block_asm_host_order Make line 33 .type md5_block_asm_host_order , @function and it doesn't complain until line 707 A. -- Tim RiceMultitalents(707) 887-1469 [EMAIL PROTECTED] __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
On Mon, 23 Dec 2002, Andy Polyakov wrote: My solution was to add support for assembly modules. We probably have to postpone this patch to 0.9.7a. If you only could reply more swiftly so that the changes could be exposed in beta... Most improtantly I don't think so! It must be complaining about leal .LXXXcbc_jmp_table-.LXXXPIC_point(%edx),%ecx. Could you examine these lines closer? Can you confirm that leal 0+[.LXXXcbc_jmp_table-.LXXXPIC_point](%edx),%ecx does the trick? That does not seem to help. I've attached the .s file in case that will help. -- Tim RiceMultitalents(707) 887-1469 [EMAIL PROTECTED] # 1 asm/dx86unix.cpp # 21 asm/dx86unix.cpp # 32 asm/dx86unix.cpp .file des-586.s .version01.01 gcc2_compiled.: .text .align 4 .globl DES_encrypt1 DES_encrypt1: pushl %esi pushl %edi movl12(%esp), %esi xorl%ecx, %ecx pushl %ebx pushl %ebp movl(%esi), %eax movl28(%esp), %ebx movl4(%esi),%edi roll$4, %eax movl%eax, %esi xorl%edi, %eax andl$0xf0f0f0f0,%eax xorl%eax, %esi xorl%eax, %edi roll$20,%edi movl%edi, %eax xorl%esi, %edi andl$0xffff,%edi xorl%edi, %eax xorl%edi, %esi roll$14,%eax movl%eax, %edi xorl%esi, %eax andl$0x,%eax xorl%eax, %edi xorl%eax, %esi roll$22,%esi movl%esi, %eax xorl%edi, %esi andl$0x03fc03fc,%esi xorl%esi, %eax xorl%esi, %edi roll$9, %eax movl%eax, %esi xorl%edi, %eax andl$0x,%eax xorl%eax, %esi xorl%eax, %edi .byte 209 .byte 199 lealDES_SPtrans,%ebp movl24(%esp), %ecx cmpl$0, %ebx je .L000start_decrypt movl(%ecx), %eax xorl%ebx, %ebx movl4(%ecx),%edx xorl%esi, %eax xorl%ecx, %ecx xorl%esi, %edx andl$0xfcfcfcfc,%eax andl$0xcfcfcfcf,%edx movb%al,%bl movb%ah,%cl rorl$4, %edx xorl (%ebp,%ebx),%edi movb%dl,%bl xorl0x200(%ebp,%ecx),%edi movb%dh,%cl shrl$16,%eax xorl0x100(%ebp,%ebx),%edi movb%ah,%bl shrl$16,%edx xorl0x300(%ebp,%ecx),%edi movb%dh,%cl andl$0xff, %eax andl$0xff, %edx xorl0x600(%ebp,%ebx),%edi xorl0x700(%ebp,%ecx),%edi movl24(%esp), %ecx xorl0x400(%ebp,%eax),%edi xorl0x500(%ebp,%edx),%edi movl8(%ecx),%eax xorl%ebx, %ebx movl12(%ecx), %edx xorl%edi, %eax xorl%ecx, %ecx xorl%edi, %edx andl$0xfcfcfcfc,%eax andl$0xcfcfcfcf,%edx movb%al,%bl movb%ah,%cl rorl$4, %edx xorl (%ebp,%ebx),%esi movb%dl,%bl xorl0x200(%ebp,%ecx),%esi movb%dh,%cl shrl$16,%eax xorl0x100(%ebp,%ebx),%esi movb%ah,%bl shrl$16,%edx xorl0x300(%ebp,%ecx),%esi movb%dh,%cl andl$0xff, %eax andl$0xff, %edx xorl0x600(%ebp,%ebx),%esi xorl0x700(%ebp,%ecx),%esi movl24(%esp), %ecx xorl0x400(%ebp,%eax),%esi xorl0x500(%ebp,%edx),%esi movl16(%ecx), %eax xorl%ebx, %ebx movl20(%ecx), %edx xorl%esi,
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
On Mon, 23 Dec 2002, Andy Polyakov wrote: But in either case I was actually thinking about something like this: ... sed -e 's/ +\([\.,:@]\) +/\1/g' -e 's/#.*//' ... Doesn't work here. Ie. doesn't do what ... sed -e 's/\. /./g' -e 's/@ /@/' ... did. for the unified rule. I.e. *more* aggressive than anything proposed/implemented so far. A. -- Tim RiceMultitalents(707) 887-1469 [EMAIL PROTECTED] __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
In message [EMAIL PROTECTED] on Tue, 24 Dec 2002 09:23:47 -0800 (PST), Tim Rice [EMAIL PROTECTED] said: tim On Mon, 23 Dec 2002, Andy Polyakov wrote: tim tim But in either case I was actually thinking about something like this: tim tim ... sed -e 's/ +\([\.,:@]\) +/\1/g' -e 's/#.*//' ... tim tim Doesn't work here. tim Ie. doesn't do what ... sed -e 's/\. /./g' -e 's/@ /@/' ... did. It's probably because your sed is as lossy as mine, and doesn't grok +. For some examples you've shown before, the following gives the desired result: sed -e 's/ \?\([\.,@]\) */\1/g' -e 's/ *:/:/g' -e 's/#.*//' Note that it doesn't take away more than one leading space, because, as far as I understand, some dotted commands might otherwise fail because they end up in column 0. The leading space must be made of type '0 or 1' since two consecutive characters that need to be de-spacified wouldn't get completely processed. The colon is treated separately, as it's commonly an ending delimiter rather than an in-the-middle character, and at last the filter of comments... Results on a few lines you presented in an earlier mail (consecutive pairs of input and output): . type md5_block_asm_host_order , @ function .type md5_block_asm_host_order,@function . size md5_block_asm_host_order , . L_md5_block_asm_host_order_end - md .size md5_block_asm_host_order,.L_md5_block_asm_host_order_end - md Appro, would the above sed fit your requirements? -- Richard Levitte \ Spannvägen 38, II \ [EMAIL PROTECTED] Redakteur@Stacken \ S-168 35 BROMMA \ T: +46-8-26 52 47 \ SWEDEN \ or +46-708-26 53 44 Procurator Odiosus Ex Infernis-- [EMAIL PROTECTED] Member of the OpenSSL development team: http://www.openssl.org/ Unsolicited commercial email is subject to an archival fee of $400. See http://www.stacken.kth.se/~levitte/mail/ for more info. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
On Wed, 25 Dec 2002, Richard Levitte - VMS Whacker wrote: In message [EMAIL PROTECTED] on Tue, 24 Dec 2002 09:23:47 -0800 (PST), Tim Rice [EMAIL PROTECTED] said: tim On Mon, 23 Dec 2002, Andy Polyakov wrote: tim tim But in either case I was actually thinking about something like this: tim tim ... sed -e 's/ +\([\.,:@]\) +/\1/g' -e 's/#.*//' ... tim tim Doesn't work here. tim Ie. doesn't do what ... sed -e 's/\. /./g' -e 's/@ /@/' ... did. It's probably because your sed is as lossy as mine, and doesn't grok +. For some examples you've shown before, the following gives the desired result: sed -e 's/ \?\([\.,@]\) */\1/g' -e 's/ *:/:/g' -e 's/#.*//' It doesn't work with the sed on UnixWare or sed on SCO OpenServer. The 's/#.*//' rule works. It seems to do the right thing with GNU sed. -- Tim RiceMultitalents(707) 887-1469 [EMAIL PROTECTED] __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
My solution was to add support for assembly modules. We probably have to postpone this patch to 0.9.7a. If you only could reply more swiftly so that the changes could be exposed in beta... Most improtantly I don't think so! It must be complaining about leal .LXXXcbc_jmp_table-.LXXXPIC_point(%edx),%ecx. Could you examine these lines closer? Can you confirm that leal 0+[.LXXXcbc_jmp_table-.LXXXPIC_point](%edx),%ecx does the trick? why can't we have a unified rule? I mean something like this: asm/dx86-elf.o: asm/dx86unix.cpp $(CPP) -DELF asm/sx86unix.cpp | \ sed -e 's/\. /./g' -e 's/@ /@/' -e 's/ :/:/' -e 's/^#.*//' \ /tmp/dx86.s \ as -o asm/dx86-elf.o /tmp/dx86.s; \ rm -f /tmp/dx86.s Is it $(CPP) in question which adds them? Yes, $(CPP) adds the spaces. I can imagine / :/:/, but where do /\. /./ and /@ /@/ some from? Can you pinpoint the lines? I mean just give the offending lines' number... A. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
In message [EMAIL PROTECTED] on Mon, 23 Dec 2002 13:16:40 +0100, Andy Polyakov [EMAIL PROTECTED] said: appro I can imagine / :/:/, but where do /\. /./ and /@ /@/ some from? Can you appro pinpoint the lines? I mean just give the offending lines' number... Actually, I have no problem seeing that spaces might be added at least around '.' by $(CPP). Some modern C preprocessors do separate tokens with spaces, and '.' is considered a separate token. -- Richard Levitte \ Spannvägen 38, II \ [EMAIL PROTECTED] Redakteur@Stacken \ S-168 35 BROMMA \ T: +46-8-26 52 47 \ SWEDEN \ or +46-708-26 53 44 Procurator Odiosus Ex Infernis-- [EMAIL PROTECTED] Member of the OpenSSL development team: http://www.openssl.org/ Unsolicited commercial email is subject to an archival fee of $400. See http://www.stacken.kth.se/~levitte/mail/ for more info. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
appro I can imagine / :/:/, but where do /\. /./ and /@ /@/ some from? Can you appro pinpoint the lines? I mean just give the offending lines' number... Actually, I have no problem seeing that spaces might be added at least around '.' by $(CPP). Some modern C preprocessors do separate tokens with spaces, Even if they are not engaged in any preprocessing activity around the affected code? I mean I can perfectly understand that #define a b a.c;b.c gets preprocessed as b .c;b.c, but not as b . c ; b . c. Can you really confirm that you've observed the latter behaviour (or similar)? and '.' is considered a separate token. Not in floating point constants:-) But in either case I was actually thinking about something like this: ... sed -e 's/ +\([\.,:@]\) +/\1/g' -e 's/#.*//' ... for the unified rule. I.e. *more* aggressive than anything proposed/implemented so far. A. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
In message [EMAIL PROTECTED] on Mon, 23 Dec 2002 14:57:00 +0100, Andy Polyakov [EMAIL PROTECTED] said: appro #define a b appro a.c;b.c appro appro gets preprocessed as b .c;b.c, but not as b . c ; b . c. Can you appro really confirm that you've observed the latter behaviour (or similar)? I have observed the latter behaviour in DEC C when I worked on Emacs for VMS. This was years ago, and now I fail to reproduce it... appro and '.' is considered a separate token. appro appro Not in floating point constants:-) True... appro But in either case I was actually thinking about something like this: appro appro ... sed -e 's/ +\([\.,:@]\) +/\1/g' -e 's/#.*//' ... appro appro for the unified rule. I.e. *more* aggressive than anything appro proposed/implemented so far. It's probably a good idea to do that... -- Richard Levitte \ Spannvägen 38, II \ [EMAIL PROTECTED] Redakteur@Stacken \ S-168 35 BROMMA \ T: +46-8-26 52 47 \ SWEDEN \ or +46-708-26 53 44 Procurator Odiosus Ex Infernis-- [EMAIL PROTECTED] Member of the OpenSSL development team: http://www.openssl.org/ Unsolicited commercial email is subject to an archival fee of $400. See http://www.stacken.kth.se/~levitte/mail/ for more info. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
On Sat, 14 Dec 2002, Andy Polyakov wrote: My solution was to add support for assembly modules. Why do you want to add more rules? Well, I actually fail to understand To fix the platforms I have access to without breaking those I don't have access to. :-) why can't we have a unified rule? I mean something like this: asm/dx86-elf.o: asm/dx86unix.cpp $(CPP) -DELF asm/sx86unix.cpp | \ sed -e 's/\. /./g' -e 's/@ /@/' -e 's/ :/:/' -e 's/^#.*//' \ /tmp/dx86.s \ as -o asm/dx86-elf.o /tmp/dx86.s; \ rm -f /tmp/dx86.s Something like that should work fine for UnixWare and Solaris x86 Unfortunatly it's not enough for SCO OpenServer. OpenServer needs -b elf passwd to as which would compensate for all observed stupid things. And even for Solaris... The only thing that would justify a new rule should be another -D to be passed to $(CPP). BTW, could you comment on where are those extra spaces you have to get rid of coming from? Is it $(CPP) in question which adds them? Yes, $(CPP) adds the spaces. [snip] -- Tim RiceMultitalents(707) 887-1469 [EMAIL PROTECTED] __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
My solution was to add support for assembly modules. Why do you want to add more rules? Well, I actually fail to understand why can't we have a unified rule? I mean something like this: asm/dx86-elf.o: asm/dx86unix.cpp $(CPP) -DELF asm/sx86unix.cpp | \ sed -e 's/\. /./g' -e 's/@ /@/' -e 's/ :/:/' -e 's/^#.*//' \ /tmp/dx86.s \ as -o asm/dx86-elf.o /tmp/dx86.s; \ rm -f /tmp/dx86.s which would compensate for all observed stupid things. And even for Solaris... The only thing that would justify a new rule should be another -D to be passed to $(CPP). BTW, could you comment on where are those extra spaces you have to get rid of coming from? Is it $(CPP) in question which adds them? But now with the latest changes, without my patch same as before (SCO OpenServer build was broken, UnixWare is OK) So nothing has changed:-) When I add my patch updating for the PIC stuff in des/Makefile.ssl, Both OpenServer and UnixWare are broken. Here are the errors. Perhaps someone can point me in the right direction. (I'll figure it out sooner or later but pointer are allways welcome) sco3-gcc (in crypto/des) (cd asm; /usr/local/bin/perl des-586.pl cpp dx86unix.cpp) gcc -E -DCOFF asm/dx86unix.cpp | sed 's/ :/:/' asm/dx86coff.s as -o asm/dx86coff.o asm/dx86coff.s Assembler: des-586.s aline 2460 : Illegal subtraction aline 2647 : Illegal subtraction *** Error code 127 Looks like is doesn't like sall$8, %ecx I don't think so! It must be complaining about leal .LXXXcbc_jmp_table-.LXXXPIC_point(%edx),%ecx. Could you examine these lines closer? Attach dx86coff.s if you have to. The unixware-2.1-pentium, unixware-7-pentium, sco5-cc-pentium, sco5-gcc builds all get this error when linking apps/openssl Undefined first referenced symbol in file fcrypt_body ../libcrypto.a(fcrypt.o) DES_ede3_cbc_encryptspeed.o DES_ncbc_encryptspeed.o DES_encrypt1../libcrypto.a(ecb_enc.o) DES_encrypt3../libcrypto.a(ecb3_enc.o) DES_decrypt3../libcrypto.a(ecb3_enc.o) openssl: fatal error: Symbol referencing errors. No output written to openssl *** Error code 1 (bu21) gmake: *** [sub_all] Error 1 Well, this only means that it failed to build corresponding ./crypto/des/asm/*x86*.o and you should look up the build log to figure out why it actually failed. A. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
The crucial thing to test is that things are still working properly in Windows, especially the DES assembler modules. They been changed to generate PIC code on Unix, and it's important that we get tests on how that affects Windows, if it does. Try openssl-0.9.7-SNAP-20021214 instead as it becomes available. A. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
Yes, indeed, win32 asm (both nasm + masm) are broken in openssl-0.9.7-stable-SNAP-20021213. Nasm gives: === [NASM version 0.98.35 compiled on Oct 28 2002] nasmw -f win32 -o crypto\des\asm\d_win32.obj .\crypto\des\asm\d_win32.asm .\crypto\des\asm\d_win32.asm:60: error: expression syntax error .\crypto\des\asm\d_win32.asm:1088: error: expression syntax error NMAKE : fatal error U1077: 'nasmw' : return code '0x1' Stop. The offending lines are both: lea ebp, [_DES_SPtrans+] Masm gives: === Microsoft (R) Macro Assembler Version 6.11d Copyright (C) Microsoft Corp 1981-1995. All rights reserved. Assembling: .\crypto\des\asm\d_win32.asm .\crypto\des\asm\d_win32.asm(63) : error A2016: expression expected .\crypto\des\asm\d_win32.asm(1095) : error A2016: expression expected .\crypto\des\asm\d_win32.asm(2373) : error A2006: undefined symbol : _$L008PIC_point .\crypto\des\asm\d_win32.asm(2376) : error A2032: invalid use of register .\crypto\des\asm\d_win32.asm(2555) : error A2006: undefined symbol : _$L032PIC_point .\crypto\des\asm\d_win32.asm(2558) : error A2032: invalid use of register NMAKE : fatal error U1077: 'ml' : return code '0x1' Stop. Apparently it doesn't like the $'s. I also tried Masm 6.15 with the same results. (For completeness' sake, the ms\do_ms.bat configuration builds and tests fine.) Openssl-0.9.7-stable-SNAP-20021212 builds fine for do_nasm, do_masm and do_ms. If someone makes a patch I can try again on Monday. -Nathan Richard Levitte - VMS Whacker wrote: It will be available tomorrow. The crucial thing to test is that things are still working properly in Windows, especially the DES assembler modules. They been changed to generate PIC code on Unix, and it's important that we get tests on how that affects Windows, if it does. -- Richard Levitte \ Spannvägen 38, II \ [EMAIL PROTECTED] Redakteur@Stacken \ S-168 35 BROMMA \ T: +46-8-26 52 47 \ SWEDEN \ or +46-708-26 53 44 Procurator Odiosus Ex Infernis-- [EMAIL PROTECTED] Member of the OpenSSL development team: http://www.openssl.org/ Unsolicited commercial email is subject to an archival fee of $400. See http://www.stacken.kth.se/~levitte/mail/ for more info. __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED] __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]
Re: IMPORTANT: please test snapshot openssl-0.9.7-SNAP-20021213
On Fri, 13 Dec 2002, Richard Levitte - VMS Whacker wrote: It will be available tomorrow. The crucial thing to test is that things are still working properly in Windows, especially the DES assembler modules. They been changed to generate PIC code on Unix, and it's important that we get tests on how that affects Windows, if it does. The changes affect some UNIX platforms too. Before the changes SCO OpenServer build was broken. (UnixWare was OK) sco3-gcc sco5-cc-pentium would get this error when building apps/openssl ... undefined first referenced symbol in file bn_sub_part_words ../libcrypto.a ... sco5-gcc would bomb in crypto/md5 with ... making all in crypto/md5... gcc -I.. -I../.. -I../../include -DDSO_DLFCN -DHAVE_DLFCN_H -DOPENSSL_NO_KRB5 -DOPENSSL_NO_IDEA -DOPENSSL_NO_RC5 -DOPENSSL_NO_MDC2 -DOPENSSL_NO_MD2 -O3 -fomit-frame-pointer -DSHA1_ASM -DMD5_ASM -DRMD160_ASM -c md5_dgst.c gcc -I.. -I../.. -I../../include -DDSO_DLFCN -DHAVE_DLFCN_H -DOPENSSL_NO_KRB5 -DOPENSSL_NO_IDEA -DOPENSSL_NO_RC5 -DOPENSSL_NO_MDC2 -DOPENSSL_NO_MD2 -O3 -fomit-frame-pointer -DSHA1_ASM -DMD5_ASM -DRMD160_ASM -c md5_one.c (cd asm; /usr/bin/perl md5-586.pl cpp mx86unix.cpp) gcc -E -DELF -x c asm/mx86unix.cpp | as -o asm/mx86-elf.o Nonexistent file cpp: output pipe has been closed *** Error code 1 (bu21) ... My solution was to add support for assembly modules. And while I was at it, I added assembly support for UnixWare. See e-mail to openssl-dev@ Date: Mon, 9 Dec 2002 16:19:00 -0800 (PST) From: Tim Rice [EMAIL PROTECTED] To: Cryptography notification [EMAIL PROTECTED], [EMAIL PROTECTED] Cc: [EMAIL PROTECTED] Subject: TSU NOTIFICATION - encryption That patch wokrked fine and didn't break anything else. But now with the latest changes, without my patch same as before (SCO OpenServer build was broken, UnixWare is OK) When I add my patch updating for the PIC stuff in des/Makefile.ssl, Both OpenServer and UnixWare are broken. Here are the errors. Perhaps someone can point me in the right direction. (I'll figure it out sooner or later but pointer are allways welcome) sco3-gcc (in crypto/des) (cd asm; /usr/local/bin/perl des-586.pl cpp dx86unix.cpp) gcc -E -DCOFF asm/dx86unix.cpp | sed 's/ :/:/' asm/dx86coff.s as -o asm/dx86coff.o asm/dx86coff.s Assembler: des-586.s aline 2460 : Illegal subtraction aline 2647 : Illegal subtraction *** Error code 127 Looks like is doesn't like sall$8, %ecx The unixware-2.1-pentium, unixware-7-pentium, sco5-cc-pentium, sco5-gcc builds all get this error when linking apps/openssl Undefined first referenced symbol in file fcrypt_body ../libcrypto.a(fcrypt.o) DES_ede3_cbc_encryptspeed.o DES_ncbc_encryptspeed.o DES_encrypt1../libcrypto.a(ecb_enc.o) DES_encrypt3../libcrypto.a(ecb3_enc.o) DES_decrypt3../libcrypto.a(ecb3_enc.o) openssl: fatal error: Symbol referencing errors. No output written to openssl *** Error code 1 (bu21) gmake: *** [sub_all] Error 1 I'm out of time today (friday). I'll try and get back to it Sunday. -- Tim RiceMultitalents(707) 887-1469 [EMAIL PROTECTED] __ OpenSSL Project http://www.openssl.org Development Mailing List [EMAIL PROTECTED] Automated List Manager [EMAIL PROTECTED]