cabs2cil: don't let extern inline prototypes replace definitions of… #25
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…he same function seen earlier.
|
Can you please add a test? If you are lost in CIL's test infrastructure, please just give a minimal C program that would compile (and/or run) with your patch, but not without. |
|
Ah, looks like you already did that (a long time ago) in issue #14. I'll try to re-use your program from there (but it will likely be faster if you update this PR with it). Thanks a lot for the patch anyway :-) |
Previously CIL's handling of gcc's alias attribute was broken. It would create a new function, rather than a linker alias, and would not work for varargs functions or any kind of variable. This patch removes that code in favour of a "do nothing" implementation which correctly passes the simple test case provided.
… fix in commit bf49a9).
When a static local is promoted to a global, the alpha renaming usually remembers any suffix added and chooses a fresh suffix for later vars of the same name. However, if the same function also has a (non-static) local with the same name, this gets clobbered: the alpha renamer throws away the renaming state after the function scope is done. This can be triggered if (among other cases) two functions in the same file both have like-named statics hiding other locals/formals (seen in Motif).
|
Once again my git skills have let me down... I've pushed a new commit to my fork, adding the test case that you (very reasonably) asked for. I've also pushed two unrelated commits that belong in separate pull requests. I've no idea how to exclude those two from this pull request... maybe you can cherry-pick? |
|
git rebase -i origin/develop followed by git push -f is generally how
you solve that situation, but I'll cherry pick, don't worry. Thanks
again for you help.
|
…stdin and stdout. Previously single-character output filenames, including "-" for stdout, were being substituted with generated names. Among other things, this broke the assemble-to-stdout pattern used in some configure scripts. Similarly, single-character input filenames were not properly supported. We make the following small changes and also add a test case. Make cilly accept input on stdin, and understand gcc's "-x <lang>" option OutputFile.pm: prevent output filenames that begin with a hyphen from being confused with command-line switches. Cilly.pm.in: be consistent about using stderr for diagnostic and debug output. main.ml: permit '-' as an input file argument on the command line. OutputFile.pm: permit '-' as an output filename, denoting stdout... meaning we must neither rewrite it to './-' nor do the protect check on it. Add a test case for input and output filenames of '-'.
Previously, if the user defined a function, say strstr(), which also has a builtin version, __builtin_strstr(), but the file contained calls only to the builtin version, the user's version would be ignored. This is a divergence from gcc's behaviour, where the user's function would be called instead of the builtin (even for calls directly to the builtin). The divergence is caused by CIL deleting the apparently unused user function. Fix it by marking the user-defined function as referenced whenever the builtin is marked.
…ich will make subsequent fixes easier.
Previously CIL was incapable of faithfully preserving inline semantics in all cases, because it lost information when merging declarations and definitions into a single varinfo. We now remember all that information, and ensure when printing that we generate adequate prototypes to represent the distinct cases under both C99's rules and GNU C's rules governing "inline" and "extern inline".
Previously the parser would reject extended-form inline assembly statements with clobber list entries consisting of multiple implicitly concatenated string literals. This caused parse errors in some code, such as in ffmpeg, that uses heavily macroized inline assembly.
…velop Conflicts: src/frontc/cabs2cil.ml
|
I'm closing this since I have recently added a bunch more changes on my fork. They are all ready for pulling, but I'll request them one-by-one, rather than this request which is about pulling my whole develop/ branch in one go. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
…the same function seen earlier.
Currently, if a .c file defines an extern inline function and then repeats its prototype without a definition, the definition is omitted from the .cil.c output file. This tiny patch fixes that bug.