Re: any use to build from source?
Chuck Swiger <[EMAIL PROTECTED]> writes: > One way is to put your local changes into files/patch-aa [1] using > diff format. Other times it's as simple as defining some environment > variables by passing them into make, via /etc/make.conf, etc. But what happens to the file files/patch-aa after I do cd /usr/src make update ? (I use CVSup to keep current.) I'm sorry for not mentioning this at the beginning. Kai ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: any use to build from source?
Kai Grossjohann wrote: Charles Swiger <[EMAIL PROTECTED]> writes: Oh, yes. The first time you run into a problem and fix it yourself, or make a change to the programs to add some feature that you want, you will discover the serious advantages. However, if you never try to fix bugs or write code for yourself, then you aren't going to gain nearly as much from using source compared with using precompiled binaries. How does one deal with local changes in the software when installing as a port? One way is to put your local changes into files/patch-aa [1] using diff format. Other times it's as simple as defining some environment variables by passing them into make, via /etc/make.conf, etc. -- -Chuck [1]: Choose whatever name seems appropriate, perhaps files/patch-src-file.c; the patch-aa naming convention works fine but is depricated. ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: any use to build from source?
Charles Swiger <[EMAIL PROTECTED]> writes: > Oh, yes. The first time you run into a problem and fix it yourself, > or make a change to the programs to add some feature that you want, > you will discover the serious advantages. > > However, if you never try to fix bugs or write code for yourself, then > you aren't going to gain nearly as much from using source compared > with using precompiled binaries. How does one deal with local changes in the software when installing as a port? Kai ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: any use to build from source?
On the other hand, the OpenBSD-people advise using packages instead of ports. See http://www.openbsd.org/faq/faq8.html#PortsvsPkgs I guess it's just a matter of personal taste and needs. GH On Sat, Jun 19, 2004 at 05:11:22PM +0300, Giorgos Keramidas wrote: > On 2004-06-19 10:58, Patrick Useldinger <[EMAIL PROTECTED]> wrote: > > Giorgos Keramidas wrote: > > > > > Before I answer to this question, I cannot help noting that you don't > > > *HAVE* to compile everything from source. In fact, if you install a > > > RELEASE version of FreeBSD and use pkg_add to install the binary, > > > precompiled packages of just the applications you are going to > > > use... there is absolutely no need to rebuild anything from source. > > > > True for the CDs. But once you want to upgrade, things get more > > complicated. For example, I did not find a package for OpenOffice 1.1.1 > > in the "offical" places, although OO is certainly an excellent candidate > > for a package. > > Indeed, packages-4-stable, packages-4.10-release on ftp.freebsd.org > don't include openoffice. A search at google though yields: > > http://projects.imp.ch/openoffice/ > > which does list FreeBSD packages of OO-1.0.3 and OO-1.1.0 :-) > > > This led me to the conclusion that packages, in the FBSD world, are > > considered less important than the very well maintained ports. > > The ports people are going through a lot of effort to build, test and > package the maximum possible number of ports before each release. > > > I would prefer it to be the other way round: go for packages, unless you > > want to tweak anything. > > This is preferable from the end-user's perspective, but I think it would > exponentially increase the number of precompiled binaries the mirrors > would have to keep available. If a port has 3 options and depends on > another with 4 options, to host every possible combination that one > might want on the FTP site 12 different combinations would have to be > built and packaged! With thousands of ports in the tree this means a > mind-boggingly huge number of different builds and packages can be built. > > Is it possible to satisfy all the users with precompiled packages? No. > > So, going for ports and using packages only is a bit difficult :-/ > > > I do not agree with an earlier argument, which was that you could change > > the source. I have been programming for 25 years now, I am certain that > > you don't change code, not even in a reasonably sized project, without > > spending a large amount of time. > > I have a local patch to fetchmail in my local /usr/ports tree that fixes > a bug recent versions have with APOP: > > : [EMAIL PROTECTED]:05]/usr/ports/mail/fetchmail# cvs -q up -APd > : ? files/patch-zz::apop_bug > : [EMAIL PROTECTED]:05]/usr/ports/mail/fetchmail# > : > : [EMAIL PROTECTED]:06]/usr/ports/mail/fetchmail# cat files/patch-zz\:\:apop_bug > : --- driver.c.orig Sun Mar 28 13:29:15 2004 > : +++ driver.cSun Mar 28 13:49:42 2004 > : @@ -429,7 +429,7 @@ > : /* for POP3, we can get the size of one mail only! Unfortunately, this > : * protocol specific test cannot be done elsewhere as the protocol > : * could be "auto". */ > : - if (ctl->server.protocol == P_POP3) > : + if (ctl->server.protocol == P_POP3 || ctl->server.protocol == P_APOP || > ctl->server.protocol == P_RPOP) > : fetchsizelimit = 1; > : > : /* Time to allocate memory to store the sizes */ > > This is a serious showstopper for anyone who uses fetchmail and doesn't > want to fetch all the messages in one connection (for whatever reasons). > > It didn't take me more than 15 minutes to write, but then I'm working as > a programmer so that's "normal". Fetchmail is, IMHO, a reasonably sized > project. I'm not saying this to sound insulting to you in any way, or > to boast about my ''l33t h4x0r skillz'' -- that's nonsense. I am only > bringing it up as a good example where building the port *does* have > obvious advantages. > > - Giorgos > > ___ > [EMAIL PROTECTED] mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-questions > To unsubscribe, send any mail to "[EMAIL PROTECTED]" ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: any use to build from source?
On 2004-06-19 17:59, Patrick Useldinger <[EMAIL PROTECTED]> wrote: > Giorgos Keramidas wrote: > > > >Is it possible to satisfy all the users with precompiled packages? > >No. > > My argument is the other way round: build a package will *all* available > options. It will be bloated, but still smaller (in download size) and > faster to install. If you like it, keep it, and want to tweak it > afterwards, OK, go for it, the investment is worthwile > > That makes one single package that should suit everybody (unless options > are mutually exclusive, of course, but that's not often the case AFAIK). Heh. Not really. For instance, it wouldn't suit me. I know it sounds a bit selfish, but I mean that someone, somewhere will want their ports trimmed in size and features to suit exactly what they need and only that. Building all the features and all the possible modules of all the packages and installing all of them, at the same time, is the "Redhat, Mandrake and Fedora philosophy" that I specifically wanted to avoid when I first came to BSD. I sure hope size-bloat and feature-bloat in packages does not become the BSD standard any time soon now :( > That is indeed a reasonable example, but I am not sure there are many > of them. Maybe. I'm not the one to judge; that's for sure. ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: any use to build from source?
Giorgos Keramidas wrote: Indeed, packages-4-stable, packages-4.10-release on ftp.freebsd.org don't include openoffice. A search at google though yields: http://projects.imp.ch/openoffice/ which does list FreeBSD packages of OO-1.0.3 and OO-1.1.0 :-) Which is not 1.1.1 or the latest 1.1.2. This is preferable from the end-user's perspective, but I think it would exponentially increase the number of precompiled binaries the mirrors would have to keep available. If a port has 3 options and depends on another with 4 options, to host every possible combination that one might want on the FTP site 12 different combinations would have to be built and packaged! With thousands of ports in the tree this means a mind-boggingly huge number of different builds and packages can be built. Is it possible to satisfy all the users with precompiled packages? No. My argument is the other way round: build a package will *all* available options. It will be bloated, but still smaller (in download size) and faster to install. If you like it, keep it, and want to tweak it afterwards, OK, go for it, the investment is worthwile That makes one single package that should suit everybody (unless options are mutually exclusive, of course, but that's not often the case AFAIK). I believe that, for most software, we are just end-users. I have a local patch to fetchmail in my local /usr/ports tree that fixes a bug recent versions have with APOP: It didn't take me more than 15 minutes to write, but then I'm working as a programmer so that's "normal". Fetchmail is, IMHO, a reasonably sized project. I'm not saying this to sound insulting to you in any way, or to boast about my ''l33t h4x0r skillz'' -- that's nonsense. I am only bringing it up as a good example where building the port *does* have obvious advantages. That is indeed a reasonable example, but I am not sure there are many of them. I also believe that the original author should, if at all possible, do this correction, because it is useful to many people. -pu ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: any use to build from source?
Patrick Useldinger writes: > True for the CDs. But once you want to upgrade, things get more > complicated. For example, I did not find a package for OpenOffice > 1.1.1 in the "offical" places, although OO is certainly an > excellent candidate for a package. If you have not found in unofficial places, try here: http://projects.imp.ch/openoffice/ > This led me to the conclusion that packages, in the FBSD world, > are considered less important than the very well maintained > ports. I do not know if it is official policy, but it is de facto true for one reason. It is much easier to go from the port to the package than vice versa. > I agree with that argument, you can tailor the compilation. But > it's probably not systematical, but rather the exception. You might be surprised. With the ability to tailor things by variables fed to make (which can be made the default by inclusion in pkg_tools.conf), quite a few people (raises hand) are doing at least some of this. > I do not agree with an earlier argument, which was that you could > change the source. I have been programming for 25 years now, I am > certain that you don't change code, not even in a reasonably > sized project, without spending a large amount of time. But if the programmer and/or port maintainer do all the work for you? Robert Huff ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: any use to build from source?
On 2004-06-19 10:58, Patrick Useldinger <[EMAIL PROTECTED]> wrote: > Giorgos Keramidas wrote: > > > Before I answer to this question, I cannot help noting that you don't > > *HAVE* to compile everything from source. In fact, if you install a > > RELEASE version of FreeBSD and use pkg_add to install the binary, > > precompiled packages of just the applications you are going to > > use... there is absolutely no need to rebuild anything from source. > > True for the CDs. But once you want to upgrade, things get more > complicated. For example, I did not find a package for OpenOffice 1.1.1 > in the "offical" places, although OO is certainly an excellent candidate > for a package. Indeed, packages-4-stable, packages-4.10-release on ftp.freebsd.org don't include openoffice. A search at google though yields: http://projects.imp.ch/openoffice/ which does list FreeBSD packages of OO-1.0.3 and OO-1.1.0 :-) > This led me to the conclusion that packages, in the FBSD world, are > considered less important than the very well maintained ports. The ports people are going through a lot of effort to build, test and package the maximum possible number of ports before each release. > I would prefer it to be the other way round: go for packages, unless you > want to tweak anything. This is preferable from the end-user's perspective, but I think it would exponentially increase the number of precompiled binaries the mirrors would have to keep available. If a port has 3 options and depends on another with 4 options, to host every possible combination that one might want on the FTP site 12 different combinations would have to be built and packaged! With thousands of ports in the tree this means a mind-boggingly huge number of different builds and packages can be built. Is it possible to satisfy all the users with precompiled packages? No. So, going for ports and using packages only is a bit difficult :-/ > I do not agree with an earlier argument, which was that you could change > the source. I have been programming for 25 years now, I am certain that > you don't change code, not even in a reasonably sized project, without > spending a large amount of time. I have a local patch to fetchmail in my local /usr/ports tree that fixes a bug recent versions have with APOP: : [EMAIL PROTECTED]:05]/usr/ports/mail/fetchmail# cvs -q up -APd : ? files/patch-zz::apop_bug : [EMAIL PROTECTED]:05]/usr/ports/mail/fetchmail# : : [EMAIL PROTECTED]:06]/usr/ports/mail/fetchmail# cat files/patch-zz\:\:apop_bug : --- driver.c.orig Sun Mar 28 13:29:15 2004 : +++ driver.c Sun Mar 28 13:49:42 2004 : @@ -429,7 +429,7 @@ : /* for POP3, we can get the size of one mail only! Unfortunately, this :* protocol specific test cannot be done elsewhere as the protocol :* could be "auto". */ : - if (ctl->server.protocol == P_POP3) : + if (ctl->server.protocol == P_POP3 || ctl->server.protocol == P_APOP || ctl->server.protocol == P_RPOP) : fetchsizelimit = 1; : : /* Time to allocate memory to store the sizes */ This is a serious showstopper for anyone who uses fetchmail and doesn't want to fetch all the messages in one connection (for whatever reasons). It didn't take me more than 15 minutes to write, but then I'm working as a programmer so that's "normal". Fetchmail is, IMHO, a reasonably sized project. I'm not saying this to sound insulting to you in any way, or to boast about my ''l33t h4x0r skillz'' -- that's nonsense. I am only bringing it up as a good example where building the port *does* have obvious advantages. - Giorgos ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: any use to build from source?
Giorgos Keramidas wrote: > Before I answer to this question, I cannot help noting that you don't > *HAVE* to compile everything from source. In fact, if you install a > RELEASE version of FreeBSD and use pkg_add to install the binary, > precompiled packages of just the applications you are going to > use... there is absolutely no need to rebuild anything from source. True for the CDs. But once you want to upgrade, things get more complicated. For example, I did not find a package for OpenOffice 1.1.1 in the "offical" places, although OO is certainly an excellent candidate for a package. This led me to the conclusion that packages, in the FBSD world, are considered less important than the very well maintained ports. I would prefer it to be the other way round: go for packages, unless you want to tweak anything. > Now, some of us -- actually, I feel that this is a large percentage of > the FreeBSD users, if the amount of questions posted here on this list > is of any significance at all -- a great percentage of us likes trimming > our installations; we like building our packages with the exact options > and feature sets that *we* prefer. In such cases, having the ability to > build from source is absolutely marvelous. I agree with that argument, you can tailor the compilation. But it's probably not systematical, but rather the exception. I do not agree with an earlier argument, which was that you could change the source. I have been programming for 25 years now, I am certain that you don't change code, not even in a reasonably sized project, without spending a large amount of time. -pu ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: any use to build from source?
Hi, When I was using 4.1 release, removing just the hardware that I was not using and building the Kernel gave me mutch better performance. I Dont know about internals so, probabily, this is not true anymore. Anyway, I could mount a DNS Server on a Pentium 75 Mhz with 16MB, because the GENERIC Kernel was using about 4MB and the compiled one 2,5MB. Samba had a much better performance too... That's about the Kernel, when I install any serious application I always compile from source, so I get the last stable release and no one from the Ports collection, beside the fact that normaly I have to turn on/off some options. > Hi all, > > I must say that I was initially interested in the idea of building > software from source - but I am kind of loosing it. > > Certainly, it allows you to compile with the compiler options you want, > you are able to optimize the binaries for your CPU, but: does it really > matter? Are the speed improvements really visible? > > Dependencies was another argument: you compile with the correct headers > of dependant files, well... is that really so? If you upgraded the > dependant binaries, wouldn't you get the same effect? > > One certain drawback of compiling from source is the compilation time. > Large packages like KDE or OpenOffice take ages, so you can't just > "quickly" upgrade a whole system, or a large part of it. I might add > that I am more the typical desktop user, not using my machines for real > and specific server apps. > > So, my question is basically: did you, in your experience, find that > compiling from source *really* has any serious advantages that make up > for the time it takes? > > -PU > > ___ > [EMAIL PROTECTED] mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-questions > To unsubscribe, send any mail to "[EMAIL PROTECTED]" > ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: any use to build from source?
Optimization and speed are indeed an arguable advantage of compiling from source, but another GREAT advantage is the possibility of setting compile-time options and dependencies. I.e. lots of packages have options which can be enabled/disabled only at compile time. For example whether you want Vim with the GUI-version or not (in that case it would have a lot more dependencies, such as X and GTK). Of course the package maintainers try to choose useful defaults, but they may not suite *your* specific needs. Also the version of binary packages may be a little behind on the ports. So if you want Firefox 0.9 TODAY, you'd have to compile it yourself. I compile most of my programs from source, just for fun. GH On Fri, Jun 18, 2004 at 08:29:12PM +0200, Patrick Useldinger wrote: > Hi all, > > I must say that I was initially interested in the idea of building > software from source - but I am kind of loosing it. > > Certainly, it allows you to compile with the compiler options you want, > you are able to optimize the binaries for your CPU, but: does it really > matter? Are the speed improvements really visible? > > Dependencies was another argument: you compile with the correct headers > of dependant files, well... is that really so? If you upgraded the > dependant binaries, wouldn't you get the same effect? > > One certain drawback of compiling from source is the compilation time. > Large packages like KDE or OpenOffice take ages, so you can't just > "quickly" upgrade a whole system, or a large part of it. I might add > that I am more the typical desktop user, not using my machines for real > and specific server apps. > > So, my question is basically: did you, in your experience, find that > compiling from source *really* has any serious advantages that make up > for the time it takes? > > -PU > ___ > [EMAIL PROTECTED] mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-questions > To unsubscribe, send any mail to "[EMAIL PROTECTED]" ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: any use to build from source?
On Jun 18, 2004, at 2:29 PM, Patrick Useldinger wrote: Certainly, it allows you to compile with the compiler options you want, you are able to optimize the binaries for your CPU, but: does it really matter? Are the speed improvements really visible? Tweaking the compiler flags and targetting your specific CPU architecture generally doesn't make a visible difference. More precisely, if you aren't benchmarking the changes, they are likely to not be significant enough that you can tell. However, for some categories of software, such as multimedia players and the like which want to take advantage of MMX/SSE/3Dnow!/etc, doing so can make quite a difference. Dependencies was another argument: you compile with the correct headers of dependant files, well... is that really so? If you upgraded the dependant binaries, wouldn't you get the same effect? If the dependencies are done correctly, you'll get the same effect. However, as soon as you start considering options like which version of BerkeleyDB to use, or which version of SASL, etc, the number of combinations becomes large enough that the generic precompiled packages may not correspond to your specific environment. One certain drawback of compiling from source is the compilation time. Large packages like KDE or OpenOffice take ages, so you can't just "quickly" upgrade a whole system, or a large part of it. I might add that I am more the typical desktop user, not using my machines for real and specific server apps. Well, yes. KDE and OpenOffice are vast, bloated, suffer from creeping featurism, and thus resemble the Windows products they attempt to emulate in more ways then their authors probably want them to. One could pick on Mozilla or many other large, cross-platform software packages for this sort of thing as well. Mozilla seems to internally re-implement most of the BSD sockets APIs internally, as well as bits and pieces of many other Unix/POSIX APIs. So, my question is basically: did you, in your experience, find that compiling from source *really* has any serious advantages that make up for the time it takes? Oh, yes. The first time you run into a problem and fix it yourself, or make a change to the programs to add some feature that you want, you will discover the serious advantages. However, if you never try to fix bugs or write code for yourself, then you aren't going to gain nearly as much from using source compared with using precompiled binaries. -- -Chuck ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: any use to build from source?
Parv <[EMAIL PROTECTED]> wrote: > in message <[EMAIL PROTECTED]>, > wrote Patrick Useldinger thusly... > > > > (Building from source) allows you to compile with the compiler > > options you want, you are able to optimize the binaries for your > > CPU, but: does it really matter? Are the speed improvements really > > visible? > > I haven't run any benchmarks related to speed for use of software on > a personal, non-server computer. http://www.potentialtech.com/wmoran/source.php It's not comprehensive, and I haven't had an opportunity to do the additional tests I wanted to do, but it is _some_ testing ... -- Bill Moran Potential Technologies http://www.potentialtech.com ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: any use to build from source?
On 2004-06-18 20:38, Patrick Useldinger <[EMAIL PROTECTED]> wrote: > > So, my question is basically: did you, in your experience, find that > compiling from source *really* has any serious advantages that make up > for the time it takes? Before I answer to this question, I cannot help noting that you don't *HAVE* to compile everything from source. In fact, if you install a RELEASE version of FreeBSD and use pkg_add to install the binary, precompiled packages of just the applications you are going to use... there is absolutely no need to rebuild anything from source. Well, at least, you are not obliged to. Now, some of us -- actually, I feel that this is a large percentage of the FreeBSD users, if the amount of questions posted here on this list is of any significance at all -- a great percentage of us likes trimming our installations; we like building our packages with the exact options and feature sets that *we* prefer. In such cases, having the ability to build from source is absolutely marvelous. IMHO, compiling a program to include _exactly_ the parts that you want it to have is incomparably better than loading up a large 'framework' and hooking into it with prebuilt modules at runtime. - Giorgos ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: any use to build from source?
in message <[EMAIL PROTECTED]>, wrote Patrick Useldinger thusly... > > (Building from source) allows you to compile with the compiler > options you want, you are able to optimize the binaries for your > CPU, but: does it really matter? Are the speed improvements really > visible? I haven't run any benchmarks related to speed for use of software on a personal, non-server computer. In case when debugging support in a software would be needed, already built software would be lacking. > Dependencies was another argument: you compile with the correct > headers of dependant files, well... is that really so? If you > upgraded the dependant binaries, wouldn't you get the same effect? I see compiling from source as a way to reduce and/or alter dependencies, and number/size of installed files. One drawback of using pkg_add to install a binary package is that it requires unnecessary upgrade of dependencies. If a package to be installed requires x-1.2.3 but x-1.2.2_p9 is already installed AND there will be no problem of using x-1.2.2_p9, x-1.2.3 will be installed regardless in addition to x-1.2.2_p9, quite possibly overwriting the files. The packages are almost never suitable for me even though i do not run servers other than for personal needs. > One certain drawback of compiling from source is the compilation > time. Large packages like KDE or OpenOffice take ages, so you > can't just "quickly" upgrade a whole system That sure is true, in addition to requirement of large build space too, jdk-1.4 requires ~1.7 GB for example. BTW, your concerns have had been dealt w/ in past, at least once this year. - Parv -- ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"