It has function createMCSubtargetInfo in Target class that creates subtarget for processor and its features. Unless you pass "help" string to it instead of CPU model.
In this case this function prints everything to stderr and calls exit(1), shutting down your whole application.
You write GUI program that has no stderr? Well, sucks to be you.
You can't get list of processors or features otherwise. Only by passing "help" and quiting application.
Thanks goodness LLVM has a nice license.
>>2
Typical unix retard "hurr, durr, text output is good and parsable". It isn't. Function that creates targets should not print output and terminate program. It's called "createMCSubtargetInfo", not "createMCSubtargetInfoPrintHelpAndExitWithoutCleanup".
Only complete moron could thing otherwise.
>>3
But I want to rant and vent. I already patched local version.
>>5
It's because of Unix retards like you that modern computing sucks. Guess what? ASCII text isn't the best representation for everything. Sometimes records, arrays, lists, hash tables, pictures, numbers, non-ASCII text or even a combination of these are desirable.
Go back to masturbating to your hacked up RFC ``specifications''.
Name:
Anonymous2012-07-22 5:21
>>7
ASCII text is the best representation for everything. Binary blobs can be encoded with base64. COME AT ME BRO.
and ascii is good for somethings, like help messages and usage examples, which your program intentionally invoked, by passing "help" in somewhere that triggered a help message.
Name:
Anonymous2012-07-22 5:24
>>7
Text IS the best when you want to be able to communicate with a bunch of different programs and services.
It's human readable - so easy to detect errors and debug.
It's universal and easy to parse - don't have to look up some shitty documentation for the shitty binary format.
Name:
Anonymous2012-07-22 5:27
like I cant believe people are still using C++ when they could be optimizing their code with performance aware CYTHON
we are talking about a help message that is embedded in the binary here. This is a bare minimum sort of thing that can help keep you straight on invocations if you have no access to the documentation. Images and links are easy to represent in HTML, which can provide the actual documentation.
>>14
EZ. <blob length="[i]Length of the blob[/i]">[i]The actual unaltered blob: no XML escaping, no base64[/i]</blob>
The closing tag isn't actually necessary and XML is shit though.
Name:
Anonymous2012-07-22 6:55
>>17
That's not valid XML, niggershit. ASCII and Base64 are the way!
>>4,7-18
The real WTF is that the function does something completely different and unexpected from what it should be doing, triggered by an odd use of parameter. OP would've been as irritated if passing "help" caused it to speak the documentation through the default audio device.
yeah, it should be moved into the main function of the command line implementation, or into usage functions used only by this main function. That certainly doesn't belong in a library.
Formally, assuming the axiom of choice, the cardinality of a set X is the least ordinal α such that there is a bijection between X and α. This definition is known as the von Neumann cardinal assignment. If the axiom of choice is not assumed we need to do something different. The oldest definition of the cardinality of a set X (implicit in Cantor and explicit in Frege and Principia Mathematica) is as the class [X] of all sets that are equinumerous with X. This does not work in ZFC or other related systems of axiomatic set theory because if X is non-empty, this collection is too large to be a set. In fact, for X ≠ ∅ there is an injection from the universe into [X] by mapping a set m to {m} × X and so by limitation of size, [X] is a proper class. The definition does work however in type theory and in New Foundations and related systems. However, if we restrict from this class to those equinumerous with X that have the least rank, then it will work (this is a trick due to Dana Scott: it works because the collection of objects with any given rank is a set).
>>7
The Unix approach worked well for the predominantly text oriented tasks for that the system was created to handle. There are enough compromises in the original Unix systems to demonstrate that its creators knew byte I/O wasn't the best approach for everything (there are a lot more syscalls in AT&T Unix than just read and write, after all).
The religious supposition that byte streams are the best way to handle absolutely everything are a later development. Plan 9 helped popularize this but again I would point out that even that system doesn't use read and write for everything (its designers were smart people, not ideologues).
Having this behavior in a shared library like Clang is just old fashioned bad design. There are no functions in unistd.h that normally operate on data but kill your whole program if a magic argument is given, because that is a plainly stupid thing to do.