From: Olivier Hainque <hainque@adacore.com>
To: Bernd Schmidt <bschmidt@redhat.com>
Cc: GCC Patches <gcc-patches@gcc.gnu.org>
Subject: Re: prevent "undef var" errors on gcc --help or --version
Date: Tue, 12 Jan 2016 16:11:00 -0000 [thread overview]
Message-ID: <4776E0F5-184D-44AB-83D5-E80DE8A3B323@adacore.com> (raw)
In-Reply-To: <5693D3C3.4010908@redhat.com>
[-- Attachment #1: Type: text/plain, Size: 1446 bytes --]
Hello Bernd,
Thanks for your feedback on this :-)
> On 11 Jan 2016, at 17:09, Bernd Schmidt <bschmidt@redhat.com> wrote:
>
> On 01/08/2016 02:23 PM, Olivier Hainque wrote:
>> + /* Undefined variable references in specs are harmless if
>> + we're running for --help or --version alone, or together. */
>> + spec_undefvar_allowed =
>> + (((print_version || print_help_list)
>> + && decoded_options_count == 2)
>> + ||
>> + ((print_version && print_help_list)
>> + && decoded_options_count == 3));
>> +
>
> This doesn't follow the formatting rules.
Arg, indeed. Revised version attached.
> Also, there are a couple of other options that cause gcc to just print something and exit. Are these affected by missing env vars?
Some of these, for sure. For example, a common use case here is to
define a default --sysroot. We need this to be set properly for at
least --print-search-dirs and --print-prog-name, probably --print-file-name.
The print-multi family might be ok. It's heavily based on the presence
of other options on the command line, but maybe never depending on argument
values. I wasn't ready to bet though and opted for a conservative approach
first.
The attached patch is doing the same as the previous one, except more
explicitly and making it easier to adapt if deemed useful.
I could extract the decision code in a separate function if you prefer.
Olivier
[-- Attachment #2: spec-undef.diff --]
[-- Type: application/octet-stream, Size: 2763 bytes --]
diff --git a/gcc/gcc.c b/gcc/gcc.c
index 319a073..92fb051 100644
--- a/gcc/gcc.c
+++ b/gcc/gcc.c
@@ -3299,6 +3299,11 @@ int n_infiles;
static int n_infiles_alloc;
+/* True if undefined environment variables encountered during spec processing
+ are ok to ignore, typically when we're running for --help or --version. */
+
+static bool spec_undefvar_allowed;
+
/* True if multiple input files are being compiled to a single
assembly file. */
@@ -4542,6 +4547,27 @@ process_command (unsigned int decoded_options_count,
add_infile ("help-dummy", "c");
}
+ /* Decide if undefined variable references are allowed in specs. */
+ {
+ /* --version and --help alone or together are safe. Note that -v would
+ make them unsafe, as they'd then be run for subprocesses as well, the
+ location of which might depend on variables possibly coming from
+ self-specs. */
+
+ /* Count the number of options we have for which undefined variables
+ are harmless for sure, and check that nothing else is set. */
+
+ unsigned n_varsafe_options = 0;
+
+ if (print_version)
+ n_varsafe_options++;
+
+ if (print_help_list)
+ n_varsafe_options++;
+
+ spec_undefvar_allowed = (decoded_options_count == n_varsafe_options + 1);
+ }
+
alloc_switch ();
switches[n_switches].part1 = 0;
alloc_infile ();
@@ -9085,14 +9111,17 @@ print_multilib_info (void)
\f
/* getenv built-in spec function.
- Returns the value of the environment variable given by its first
- argument, concatenated with the second argument. If the
- environment variable is not defined, a fatal error is issued. */
+ Returns the value of the environment variable given by its first argument,
+ concatenated with the second argument. If the variable is not defined, a
+ fatal error is issued unless such undefs are internally allowed, in which
+ case the variable name is used as the variable value. */
static const char *
getenv_spec_function (int argc, const char **argv)
{
const char *value;
+ const char *varname;
+
char *result;
char *ptr;
size_t len;
@@ -9100,10 +9129,15 @@ getenv_spec_function (int argc, const char **argv)
if (argc != 2)
return NULL;
- value = env.get (argv[0]);
+ varname = argv[0];
+ value = env.get (varname);
+
+ if (!value && spec_undefvar_allowed)
+ value = varname;
+
if (!value)
fatal_error (input_location,
- "environment variable %qs not defined", argv[0]);
+ "environment variable %qs not defined", varname);
/* We have to escape every character of the environment variable so
they are not interpreted as active spec characters. A
[-- Attachment #3: Type: text/plain, Size: 2 bytes --]
next prev parent reply other threads:[~2016-01-12 16:11 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-01-08 13:23 Olivier Hainque
2016-01-11 16:09 ` Bernd Schmidt
2016-01-12 16:11 ` Olivier Hainque [this message]
2016-01-12 16:14 ` Bernd Schmidt
2016-01-12 16:37 ` Olivier Hainque
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4776E0F5-184D-44AB-83D5-E80DE8A3B323@adacore.com \
--to=hainque@adacore.com \
--cc=bschmidt@redhat.com \
--cc=gcc-patches@gcc.gnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).