git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [Bug] Compat objects not added to CLAR_TEST_PROG

From
Patrick Steinhardt <ps@pks.im>
Date
Sep 9, 2025, 07:45 UTC
Message-ID
<aL_bAxZubXMOGWsu@pks.im>
In-Reply-To
<20250905213708.GB612697@coredump.intra.peff.net>
On Fri, Sep 05, 2025 at 05:37:08PM -0400, Jeff King wrote:
Show 22 quoted lines
> On Fri, Sep 05, 2025 at 03:19:50PM +0200, Osipov, Michael (IN IT IN) wrote:
> > diff -u -ur t/unit-tests/clar/clar/sandbox.h git-2.51.0.patched/t/unit-tests/clar/clar/sandbox.h
> > --- t/unit-tests/clar/clar/sandbox.h	2025-08-18 02:35:38 +0200
> > +++ t/unit-tests/clar/clar/sandbox.h	2025-09-05 14:10:52 +0200
> > @@ -2,6 +2,8 @@
> >  #include <sys/syslimits.h>
> >  #endif
> >  
> > +#include "../../../../compat/posix.h"
> > +
> >  static char _clar_path[4096 + 1];
> >  
> >  static int
> 
> ...seems like an obvious improvement. If we are compiling any C code,
> we'd want our compatibility macros, etc. Although it does get a little
> funny, as the contents of clar/ are imported from elsewhere, and now
> we're modifying that.
> 
> It looks like clar tries to handle portability on its own, so I guess
> another route is for it to add its own mkdtemp wrapper, and we'd import
> that fixed version. But it really feels like we're duplicating effort.

We're duplicating effort indeed, but that effort benefits other projects that use clar.

In any case, we already have logic to detect whether or not the platform should have `mkdtemp()`:

    #if defined(__MINGW32__)
        if (_mktemp(_clar_tempdir) == NULL)
            return -1;
        if (mkdir(_clar_tempdir, 0700) != 0)
            return -1;
    #elif defined(_WIN32)
        if (_mktemp_s(_clar_tempdir, sizeof(_clar_tempdir)) != 0)
            return -1;
        if (mkdir(_clar_tempdir, 0700) != 0)
            return -1;
    #elif defined(__sun) || defined(__TANDEM)
        if (mktemp(_clar_tempdir) == NULL)
            return -1;
        if (mkdir(_clar_tempdir, 0700) != 0)
            return -1;
    #else
        if (mkdtemp(_clar_tempdir) == NULL)
            return -1;
    #endif

So that raises the question whether HP-UX has mktemp(3p) -- if so, we can probably fix the issue like this:

diff --git a/clar/sandbox.h b/clar/sandbox.h
index ff43159..5af36f3 100644
--- a/clar/sandbox.h
+++ b/clar/sandbox.h
@@ -164,7 +164,7 @@ static int build_tempdir_path(void)
 
 	if (mkdir(_clar_tempdir, 0700) != 0)
 		return -1;
-#elif defined(__sun) || defined(__TANDEM)
+#elif defined(__sun) || defined(__TANDEM) || defined(__HPUX)
 	if (mktemp(_clar_tempdir) == NULL)
 		return -1;
 
The `__HPUX` define is pulled out of thin air, I have no idea what
preprocessor macro that system sets. But something in that spirit may
fix that issue. If so, I'm happy to fix this upstream and then pull
the latest version into Git.

Patrick
Previous: Osipov, Michael (IN IT IN)Next: Osipov, Michael (IN IT IN)
Message 4 of 7 in “[Bug] Compat objects not added to CLAR_TEST_PROG”
  1. Osipov, Michael (IN IT IN)Sep 5, 2025
  2. Jeff KingSep 5, 2025
  3. Osipov, Michael (IN IT IN)Sep 5, 2025
  4. Patrick SteinhardtSep 9, 2025
  5. Osipov, Michael (IN IT IN)Sep 9, 2025
  6. Patrick SteinhardtSep 9, 2025
  7. Osipov, Michael (IN IT IN)Sep 9, 2025

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.