510175 - Shouldn't be parsing error message to determine plugin support in gcc 4.5

Status: RESOLVED FIXED

(Developer Infrastructure :: Source Code Analysis, defect)

Description

[Arpad Borsos [:Swatinem]]

Attachment: patch

Using LANG=C in subprocesses makes output matching work no matter what the
console language is.

Alternatively you could pull this patch from http://hg.swatinem.de/dehydra/

Comment 1

[(dormant account)]

Comment on attachment 394244 [details] [diff] [review]
patch

>+    env = os.environ
>+    env['LANG'] = 'C'

good idea on LANG=C, i didn't even consider localization problems. Unfortunately I can't take the Popen changes due to having a few users on older RHEL distros. The code originally used subprocess, but hard to be uglified to work with older python.

I should add a comment about that.

Comment 2

[(dormant account)]

Correct way to fix this for 4.5 is: try_gcc_plugin_headers should superseed try_enable_treehydra because if it succeeded, we have plugin support, so no need to test for it.

Comment 3

[(dormant account)]

Attachment: configure fix

Comment 4

[(dormant account)]

http://hg.mozilla.org/rewriting-and-analysis/dehydra/rev/18bb42fe2f5b

Comment 5

[Arpad Borsos [:Swatinem]]

Attachment: actually build dehydra [pushed: comment 6]

The assignment was missing so treehydra wasn't actually being compiled.

Comment 6

[Arpad Borsos [:Swatinem]]

Comment on attachment 410766 [details] [diff] [review]
actually build dehydra [pushed: comment 6]

http://hg.mozilla.org/rewriting-and-analysis/dehydra/rev/e21bcf163eee