From: Keith M. <kei...@to...> - 2007-03-20 12:10:43
|
Giovanni Bajo wrote, quoting me: [Re: Danny Smith's statement on why he hasn't released GCC-4.x] >>> OK. I wasn't able to find those messages by searching by either google >>> (with or without specific site:, or gmane). It might well be that I did >>> not search hard enough. Does google index that mingw-users mailing list >>> at all? That may be part of the trouble. >> >> I don't know. I found this one, with a gmane search: >> http://article.gmane.org/gmane.comp.gnu.mingw.user/20241 >> >> (It's quite picky though; I needed to search for "regressions", >> written by "danny smith" -- "regression" didn't cut it :-( ). > > Well, to find this article, you had to know that the fact that there is > no 4.x releases of GCC is related to the number of regressions. This is > not something which can be deduced by other means. (In fact, I hope you > believe me that I really did try and search infos about official GCC 4.X > releases and did not found anything). I agree that it wasn't as easy to find as I'd hoped it might be. > This said, the article speaks of generic P1 and P2 regressions, which > are valid for *all* platforms (not only MinGW). Those regressions > *notwithstanding*, FSF has released GCC. MinGW is of course free to have > a more severe policy, but I would disagree with such a choice. And of course, you are free to disagree. Personally, I would be willing to accommodate users who really cannot live without GCC-4.x, but I am not a GCC developer, and would require the services of a Release Manager to maintain and support such a release. At the moment I rely on Danny, to manage GCC releases for MinGW, and he is not yet comfortable with the notion of releasing GCC-4. > Instead, I'd understand if the MinGW release was held while awaiting > fixes for severe *mingw-only* regressions (given that FSF currently does > not currently consider mingw an important platform for its release > criteria). But then, my suggestion is to keep a wiki page with an exact > list of all the Bugzilla bugs (mingw-specific, or specifically affecting > mingw in a severe form) which are helding the release, so that everybody > can know the exact status of the release, and maybe even help out. > > Had I found such a list, I would have surely inspected it and see if I > could help out somehow. Please feel free to contribute such a Wiki page; your efforts will surely be appreciated. As a starting point, you could use the content of Danny's earlier message: | Many of the dllimport fixes in 4.2+ have been backported to 4.1.2. Two | significant exceptions | come to mind: | | 1) http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31275 (The most recent of | several reports) | | 2) http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27650 | | I don't know if these are significant for most mingw users, but they | are significant to my work, so in the absence of further information I | consider them 'serious'. | | Other issues which need to be addressed are | | 1) Throwing exception objects across dll/exe boundaries. | 2) C++ strings vs dll's | http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24196 | | These two can be fixed with further hacks to the mingw-specific | w32-shared-pointer code. That has become an unacceptable (to me) | maintenance burden. A libstdc++.dll option would make it soooo much | easier. | | 3) Inefficiencies and backtrace lossage associated with use of SJLJ | exceptions. This is a really old bug. | http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14563. | Changing to Dwarf2 exceptions is not hard. And here, Danny throws down the guantlet to *you*: | If mingw users want an "official" (=="bug-reports are a mingw project | responsibility") GCC-4.1.2 mingw binary package, please explain why you | can't wait and why someone besides yourself should spend time on it. I'll take that a step further, and throw down one of my own: I'd even be willing to host *your* GCC-4.1.2 release as a MinGW snapshot, along with all of its supporting sources, on condition that you are willing to address any support queries relating to it, which are addressed to this list. >>> While I respectfully disagree with your release policy (see the rest >>> of the mail), I would like to thank you for all your feedback. I have >>> updated again the webpage, and I hope that you like the new wording >>> more. Let me know if you still find incomplete or incorrect. >> >> Thanks. It's better now, although the wording might be construed as >> mildly inflammatory. > > I am mildly annoyed by the fact that there still not even an > *unofficial* GCC 4.X binary on the mingw.org (let alone an official > release), given that it's really hard to compile GCC, So, take up my challenge -- provide and maintain it for me. > and there are very > obscure problems like that of relocations (which only few people seem > aware of), And probably fewer still, among the MinGW user community, who even care. > and that of the MSVCR*.DLL, for which it is incredibly hard > to find a good source of information I don't understand your point here. MinGW *must* work with the basic MSVCRT.DLL, which is supplied as an integral component of all current Win32 OS releases, or is acquired by installation of IE5+; it must *never* depend on any other version supplied only with Visual Studio, (or questionably redistributed from any VS release). The *definitive* source of information on MSVCRT.DLL is MSDN. > (and I'm going to update the MinGW > wiki about it myself, trying to help out the project). Please do. Any useful information contributed to the Wiki is welcome, but take care to cite only publicly available sources. > Thus, I'll keep the text as "mildly inflammatory", as long as it does > not misrepresent MinGW's official position, and it is sufficiently clear > for readers. Well, it gives the impression that we are intransigently opposed to releasing any GCC-4.x build in the forseeable future. That simply isn't true. Our resources, all of which are provided on a strictly voluntary basis, are limited. We have one individual, (Danny), who is willing to manage GCC releases; he takes a cautious approach to that, for reasons which only he can definitively explain. As a GCC user, I have no problem with that philosophy; GCC-3.4.5 is perfectly adequate for my needs, and "if it ain't broke, don't fix it" works for me. Yes, there are some users who have asked about the possibility of a forthcoming GCC-4.x release. Do they really need it? Only they can answer that. Are they willing to step forward, put in the effort, and help us to create such a release? Again, only they can answer, but to date none has taken that step. Are we ready to release GCC-4.x, in the absence of such additional voluntary assistance? Danny has already answered that with a resounding "not yet". So, at the risk of seeming inflammatory myself, I'll repeat: you provide, maintain and support it, and I'll seriously consider hosting it; beyond that, it's a case of "put up, or shut up". Regards, Keith. |