LLVM Preparing To Raise Its Visual Studio Requirements

[phoronix]

Phoronix: LLVM Preparing To Raise Its Visual Studio Requirements

As a forewarning for anyone that might be forced to use MSVC / Visual Studio at their place of employment, LLVM developers are preparing to raise their compiler requirements for building out LLVM unless any old pre-2013 MSVC users raise their concerns...

http://www.phoronix.com/scan.php?page=news_item&px=LLVM-MSVC-2013-Requirement

[hubick]

People need to keep their shit up to date anyhow, and this is the kinda thing that will make LLVM much more pleasant for the developers to work on. A big boost in day-to-day developer happiness, without any major extra costs over the long term = win.

[Alliancemd]

Re

People need to keep their shit up to date anyhow, and this is the kinda thing that will make LLVM much more pleasant for the developers to work on. A big boost in day-to-day developer happiness, without any major extra costs over the long term = win.

Industries like Automotive and Aeronautics are huge users(combined, these 2 industries most probably collect over 1 million C and C++ developers) but both of these industries don't update just because there is an update. It's always preferred to use the time tested toolchain. You can't risk to introduce bugs in critical systems...
When you will crash with your car because a bug was introduced with a new toolchain you won't say "People need to keep their shit up to date anyhow"...

[nanonyme]

Industries like Automotive and Aeronautics are huge users(combined, these 2 industries most probably collect over 1 million C and C++ developers) but both of these industries don't update just because there is an update. It's always preferred to use the time tested toolchain. You can't risk to introduce bugs in critical systems...
When you will crash with your car because a bug was introduced with a new toolchain you won't say "People need to keep their shit up to date anyhow"...

It's not just those sectors. It's typical in software engineering new MSVS isn't taken into use for old components because making sure things are compatible is expensive. I would be surprised if it was just one or two companies that are still using 2005

[Ouroboros]

Err, you guys do realize that if stability is such a concern, they probably wouldn't use LLVM on Windows and, as you said, probably won't update to LLVM 3.7+ for a while if they do.
AFAIK, LLVM doesn't even completely build/work with MSVC right now, too.

[wizard69]

Industries like Automotive and Aeronautics are huge users(combined, these 2 industries most probably collect over 1 million C and C++ developers) but both of these industries don't update just because there is an update. It's always preferred to use the time tested toolchain. You can't risk to introduce bugs in critical systems...

That maybe one perspective but with something like LLVM you are probably better off updating to find more bugs in your software. Blaming the tool chain isn't always smart, sure bugs come up in new releases but at the same time old bugs get squashed. In the case of LLVM and its compilers the vastly improved diagnostics would offset any negatives.

When you will crash with your car because a bug was introduced with a new toolchain you won't say "People need to keep their shit up to date anyhow"...

Actually they do need to keep shit up to date because there is a higher likely hood of older software having introduced a bug you aren't aware of. In the end you have no argument and frankly it would look pretty pathetic if you where in a court of law training to explain the use of decades old compilers. In the end this attitude is more about laziness than anything else.

Beyond that this is for future LLVM releases, not he current one nor the scheduled next release. It is good planning.

[carewolf]

It's not just those sectors. It's typical in software engineering new MSVS isn't taken into use for old components because making sure things are compatible is expensive. I would be surprised if it was just one or two companies that are still using 2005

Well, that is because MSVC breaks much more than other compilers, so when upgrading you are sure to find new places it miscompiles, but also old places where you have MSVC specific hacks that needs to be updated.

[erendorn]

Guys, I think you misunderstood what this is about.
This is the requirement for building llvm. This only impacts contributors to the LLVM project or people that intend to build from source instead of using binaries directly.
The 1 million developer can continue using whatever they like.

[gamerk2]

It's not just those sectors. It's typical in software engineering new MSVS isn't taken into use for old components because making sure things are compatible is expensive. I would be surprised if it was just one or two companies that are still using 2005

*waves* Though at least we have a few of us with 2008.

Oh, and we need VC++ 6.0 because we have a few projects that simply don't compile without it. Easier to keep that around then figure out what the heck of wrong with the source code.