Why does it look for a repository in the output directory ???

[Kelly O'Hair]

Lately with JPRT, I am seeing:

bash ./configure MAKE=/opt/jprt/products/P1/gnumake3.81/bin/make --with-debug-level=release --with-target-bits=32 --with-num-cores=5 --with-boot-jdk=/opt/jprt/products/P1/jdk7u7-latest/jdk1.7.0_07 --with-cups-include=/opt/jprt/products/P1/cups1.0 --with-freetype=/opt/jprt/products/P1/freetype32-2.3.4/freetype-i586
./common/autoconf/configure: line 77: hg: command not found
Running custom generated-configure.sh

I think it is a horribly bad idea for the build system to rely on the source control system.
None of the JPRT build&test systems can be expected to have 'hg' and I have zero interest in putting 'hg' on them.
The RE team builds from plain source bundles, with no .hg/ directories.
JPRT builds from source bundles without .hg/ directories.

Every dependence that the builds have on the source control system will come back to bite you.

-kto

On Oct 22, 2012, at 7:20 AM, Magnus Ihse Bursie wrote:

> On 2012-10-22 16:11, Magnus Ihse Bursie wrote:
>> On 2012-10-22 13:48, David Holmes wrote:
>>> When I run configure the first thing I see is:
>>> 
>>> abort: no repository found in '/export/users/dh198349/build-infra/b00/se-linux-i586-client-ea' (.hg not found)!
>>> 
>>> This is my output directory, so why is the build trying to run a hg command on it?
>> 
>> This is a bad side-effect of my added robustness checks to configure, which looks for actually (in the "hg status" sense) modified files before complaining that the generated file is out of date. I didn't remember that hg needed to have a proper working directory to be able to check status, and never noticed it myself since I always ran configure in the top-level repo.
>> 
>> I'll fix it. Thank you for your report.
> Fix pushed.
> 
> /Magnus
>