diff options
Diffstat (limited to 'winsup/doc/faq-programming.xml')
-rw-r--r-- | winsup/doc/faq-programming.xml | 333 |
1 files changed, 292 insertions, 41 deletions
diff --git a/winsup/doc/faq-programming.xml b/winsup/doc/faq-programming.xml index bfc27fac7..c01a127e6 100644 --- a/winsup/doc/faq-programming.xml +++ b/winsup/doc/faq-programming.xml @@ -1,5 +1,11 @@ +<?xml version="1.0" encoding='UTF-8'?> +<!DOCTYPE article PUBLIC "-//OASIS//DTD DocBook V4.5//EN" + "http://www.oasis-open.org/docbook/xml/4.5/docbookx.dtd"> <!-- faq-programming.xml --> +<qandadiv id="faq.programming"> +<title>Programming Questions</title> + <qandaentry id="faq.programming.packages"> <question><para>How do I contribute a package?</para></question> <answer> @@ -48,6 +54,287 @@ package. Or compile with the <literal>-s</literal> option to gcc. </para> </answer></qandaentry> +<qandaentry id="faq.programming.64bitporting"> +<question><para>What do I have to look out for when porting applications to 64 bit Cygwin?</para></question> +<answer> + +<para>The Cygwin x86_64 toolchain is using the +<ulink url="http://en.wikipedia.org/wiki/LLP64#64-bit_data_models">LP64</ulink> +data model. That means, in contrast to Windows, which uses an +<ulink url="http://en.wikipedia.org/wiki/LLP64#64-bit_data_models">LLP64</ulink> +data model, sizeof(long) != sizeof(int), just as on Linux.</para> + +<para>For comparison:</para> + +<screen> + Cygwin Windows Cygwin + Linux x86_64 Linux + Windows x86_64 + i686 + +sizeof(int) 4 4 4 +sizeof(long) 4 4 8 +sizeof(size_t) 4 8 8 +sizeof(void*) 4 8 8 +</screen> + +<para>This difference can result in interesting problems, especially when +using Win32 functions, especially when using pointers to Windows +datatypes like LONG, ULONG, DWORD. Given that Windows is LLP64, all of +the aforementioned types are 4 byte in size, on 32 as well as on 64 bit +Windows, while `long' on 64 bit Cygwin is 8 bytes.</para> + +<para>Take the example ReadFile:</para> + +<screen> + ReadFile (HANDLE, LPVOID, DWORD, LPDWORD, LPOVERLAPPED); +</screen> + +<para>In the 32 bit Cygwin and Mingw environments, as well as in the 64 bit +Mingw environment, it is no problem to substitute DWORD with unsigned +long:</para> + +<screen> + unsigned long number_of_bytes_read; + [...] + ReadFile (fhdl, buf, buflen, &number_of_bytes_read, NULL); +</screen> + +<para>However, in 64 bit Cygwin, using LP64, number_of_bytes_read is 8 bytes +in size. But since ReadFile expects a pointer to a 4 byte type, the function +will only change the lower 4 bytes of number_of_bytes_read on return, while +the content of the upper 4 bytes stays undefined.</para> + +<para>Here are a few <emphasis>donts</emphasis> which should help porting +applications from the known ILP32 data model of 32 bit Cygwin, to the LP64 +data model of 64 bit Cygwin. Note that these are not Cygwin-only problems. +Many Linux applications suffered the same somewhat liberal handling of +datatypes when the AMD64 CPU was new.</para> + +<itemizedlist mark="bullet"> + +<listitem><para> +<emphasis>Don't</emphasis> mix up int and long in printf/scanf. This: + +<screen> + int i; long l; + printf ("%d %ld\n", l, i); +</screen> + +may not print what you think it should. Enable the gcc options -Wformat or +-Wall, which warn about type mismatches in printf/scanf functions. + +<note>Using -Wall (optionally with -Werror to drive the point home) makes a +lot of sense in general, not only when porting code to a new platform.</note> +</para></listitem> + +<listitem><para> +<emphasis>Don't</emphasis> mix int and long pointers. + +<screen> + long *long_ptr = (long *) &my_int; /* Uh oh! */ + *long_ptr = 42; +</screen> + +The assignment will write 8 bytes to the address of my_int. Since my_int +is only 4 bytes, <emphasis>something else</emphasis> gets randomly overwritten. +Finding this kind of bug is very hard, because you will often see a problem +which has no immediate connection to the actual bug. +</para></listitem> + +<listitem><para> +<emphasis>Don't</emphasis> mix int and pointers at all! This will +<emphasis>not</emphasis> work as expected anymore: + +<screen> + void *ptr; + printf ("Pointer value is %x\n", ptr); +</screen> + +%x denotes an int argument. The value printed by printf is a 4 byte value, +so on x86_64 the printed pointer value is missing its upper 4 bytes; the output +is very likely wrong. Use %p instead, which portable across architectures: + +<screen> + void *ptr; + printf ("Pointer value is %p\n", ptr); +</screen> +</para></listitem> + +<listitem><para> +Along the same lines <emphasis>don't</emphasis> use the type int in +pointer arithmetic. Don't cast pointers to int, don't cast pointer +differences to int, and don't store pointer differences in an int type. +Use the types <literal>intptr_t</literal>, <literal>uintptr_t</literal> +and <literal>ptrdiff_t</literal> instead, they are designed for performing +architecture-independent pointer arithmetic. +</para></listitem> + +<listitem><para> +<emphasis>Don't</emphasis> make blind assumptions about the size of a POSIX +type. For instance, <literal>time_t</literal> is 8 bytes on 64 bit Cygwin, +while it is (still, at the time of writing this) 4 bytes on 32 bit Cygwin, +since time_t is based on the type long. +</para></listitem> + +<listitem><para> +<emphasis>Don't</emphasis> use functions returning pointers without declaration. +For instance + +<screen> + printf ("Error message is: %s\n", strerror (errno)); +</screen> + +This code will <emphasis>crash</emphasis>, unless you included +<filename>string.h</filename>. The implicit rule in C is that an undeclared +function is of type int. But int is 4 byte and pointers are 8 byte, so the +string pointer given to printf is missing the upper 4 bytes. +</para></listitem> + +<listitem><para> +<emphasis>Don't</emphasis> use C base types together with Win32 functions. +Keep in mind that DWORD, LONG, ULONG are <emphasis>not</emphasis> the same +as long and unsigned long. Try to use only Win32 datatypes in conjunction +with Win32 API function calls to avoid type problems. See the above +ReadFile example. Windows functions in printf calls should be treated +carefully as well. This code is common for 32 bit code, but probably prints +the wrong value on 64 bit: + +<screen> + printf ("Error message is: %lu\n", GetLastError ()); +</screen> + +Using gcc's -Wformat option would warn about this. Casting to the requested +base type helps in this case: + +<screen> + printf ("Error message is: %lu\n", (unsigned long) GetLastError ()); +</screen> +</para></listitem> + +<listitem><para> +<emphasis>Don't</emphasis> mix Windows datatypes with POSIX type-specific +MIN/MAX values. + +<screen> + unsigned long l_max = ULONG_MAX; /* That's right. */ + ULONG w32_biggest = ULONG_MAX; /* Hey, wait! What? */ + ULONG w32_biggest = UINT_MAX; /* Ok, but borderline. */ +</screen> + +Again, keep in mind that ULONG (or DWORD) is <emphasis>not</emphasis> unsigned +long but rather unsigned int on 64 bit. +</para></listitem> + +</itemizedlist> + +</answer></qandaentry> + +<qandaentry id="faq.programming.64bitporting-fail"> +<question><para>My project doesn't build at all on 64 bit Cygwin. What's up?</para></question> +<answer> + +<para>Typically reasons for that are:</para> + +<itemizedlist mark="bullet"> + +<listitem><para><literal>__CYGWIN32__</literal> is not defined in the +64 bit toolchain. This may hit a few projects which are around since before +Y2K. Check your project for occurences of <literal>__CYGWIN32__</literal> +and change them to <literal>__CYGWIN__</literal>, which is defined in the +Cygwin toolchain since 1998, to get the same Cygwin-specific code changes done. +</para></listitem> + +<listitem><para>The project maintainers took it for granted that Cygwin is +running only on i686 CPUs and the code is making this assumption blindly. +You have to check the code for such assumptions and fix them. +</para></listitem> + +<listitem><para>The project is using autotools, the +<filename>config.sub</filename> and <filename>config.guess</filename> files +are hopelessly outdated and don't recognize +<literal>x86_64-{pc,unknown}-cygwin</literal> as valid target. Update the +project configury (cygport will do this by default) and try again. +</para></listitem> + +<listitem><para>The project uses Windows functions on Cygwin and it's suffering +from the problems described in the preceeding FAQ entry. +</para></listitem> + +</itemizedlist> + +<para>In all of this cases, please make sure to fix that upstream, or send +your patches to the upstream maintainers, so the problems get fixed for the +future.</para> + +</answer></qandaentry> + +<qandaentry id="faq.programming.64bitporting-cygwin64"> +<question><para>Why is __CYGWIN64__ not defined for 64 bit?</para></question> +<answer> + +<para>There is no <literal>__CYGWIN64__</literal> because we would like to +have a unified way to handle Cygwin code in portable projects. Using +<literal>__CYGWIN32__</literal> and <literal>__CYGWIN64__</literal> only +complicates the code for no good reason. Along the same lines you won't +find predefined macros <literal>__linux32__</literal> and +<literal>__linux64__</literal> on Linux.</para> + +<para>If you really have to differ between 32 and 64 bit in some way, you have +three choices.</para> + +<itemizedlist mark="bullet"> + +<listitem><para>If your code depends on the CPU architecture, use the +predefined compiler definition for the architecture, like this:</para> + +<screen> +#ifdef __CYGWIN__ +# ifdef __x86_64__ /* Alternatively __x86_64, __amd64__, __amd64 */ + /* Code specific for AMD64 CPU */ +# elif __X86__ + /* Code specific for ix86 CPUs */ +# else +# error Unsupported Architecture +# endif +#endif +</screen></listitem> + +<listitem><para>If your code depends on differences in the data model, you +should consider to use the <literal>__LP64__</literal> definition +instead:</para> + +<screen> +#ifdef __CYGWIN__ +# ifdef __LP64__ /* Alternatively _LP64 */ + /* Code specific for 64 bit CPUs */ +# else + /* Code specific for 32 bit CPUs */ +# endif +#endif +</screen></listitem> + +<listitem><para>If your code uses Windows functions, and some of the +functionality is 64 bit Windows-specific, use <literal>_WIN64</literal>, +which is defined on 64 bit Cygwin, as soon as you include +<filename>windows.h</filename>. This should only be used in the most +desperate of occasions, though, and <emphasis>only</emphasis> if it's +really about a difference in Windows API functionality!</para> + +<screen> +#ifdef __CYGWIN__ +# ifdef _WIN64 + /* Code specific for 64 bit Windows */ +# else + /* Code specific for 32 bit Windows */ +# endif +#endif +</screen></listitem> + +</itemizedlist> + +</answer></qandaentry> + <qandaentry id="faq.programming.glibc"> <question><para>Where is glibc?</para></question> <answer> @@ -62,7 +349,7 @@ would be difficult. <question><para>Where is Objective C?</para></question> <answer> -<para>Support for compiling Objective C is available in the <literal>gcc4-objc</literal> +<para>Support for compiling Objective C is available in the <literal>gcc{4}-objc</literal> package; resulting binaries will depend on the <literal>libobjc2</literal> package at runtime. </para> @@ -72,10 +359,8 @@ package at runtime. <question><para>Why does my make fail on Cygwin with an execvp error? </para></question> <answer> -<para>First of all, if you are using <literal>make -j[N]</literal>, then stop. It doesn't -work well. Also beware of using non-portable shell features in your -Makefiles (see tips at <ulink -url="http://cygwin.com/faq/faq.using.html#faq.using.shell-scripts" />). +<para>Beware of using non-portable shell features in your Makefiles (see tips +at <ulink url="http://cygwin.com/faq/faq.html#faq.using.shell-scripts" />). </para> <para>Errors of <literal>make: execvp: /bin/sh: Illegal Argument</literal> or <literal>make: execvp: /bin/sh: Argument list too long</literal> are often @@ -461,7 +746,7 @@ recommend trying the latest snapshot from <ulink url="http://cygwin.com/snapshots/" /> or building the DLL from CVS. </para> <para>To build a debugging version of the Cygwin DLL, you will need to follow -the instructions at <ulink url="http://cygwin.com/faq/faq-nochunks.html#faq.programming.building-cygwin" />. +the instructions at <ulink url="http://cygwin.com/faq/faq.html#faq.programming.building-cygwin" />. You can also contact the mailing list for pointers (a simple test case that demonstrates the bug is always welcome). </para> @@ -681,15 +966,6 @@ This is right <literal>gcc hello.cc -lstdc++</literal>. use <literal>struct stat</literal>. It's 64 bit aware.</para> </answer></qandaentry> -<qandaentry id="faq.programming.undeclared-functions"> -<question><para>I use a function I know is in the API, but I still get a link error.</para></question> -<answer> - -<para>The function probably isn't declared in the header files, or -the UNICODE stuff for it isn't filled in. -</para> -</answer></qandaentry> - <qandaentry id="faq.programming.libc"> <question><para>Can you make DLLs that are linked against libc ?</para></question> <answer> @@ -758,17 +1034,6 @@ data types, line numbers, local variables etc. </para> </answer></qandaentry> -<qandaentry id="faq.programming.x86-assembly"> -<question><para>Where can I find info on x86 assembly?</para></question> -<answer> - -<para>CPU reference manuals for Intel's current chips are available in -downloadable PDF form on Intel's web site: -</para> -<para><ulink url="http://developer.intel.com/">http://developer.intel.com/</ulink> -</para> -</answer></qandaentry> - <qandaentry id="faq.programming.make-scripts"> <question><para>Shell scripts aren't running properly from my makefiles?</para></question> <answer> @@ -849,18 +1114,4 @@ linker flag.</para></listitem> </orderedlist></listitem></orderedlist> </answer></qandaentry> -<qandaentry id="faq.programming.djgpp"> -<question><para>Why not use DJGPP ?</para></question> -<answer> - -<para>DJGPP is a similar idea, but for DOS instead of Win32. DJGPP uses a -"DOS extender" to provide a more reasonable operating interface for its -applications. The Cygwin toolset doesn't have to do this since all of -the applications are native WIN32. Applications compiled with the -Cygwin tools can access the Win32 API functions, so you can write -programs which use the Windows GUI. -</para> -<para>You can get more info on DJGPP by following -<ulink url="http://www.delorie.com/">http://www.delorie.com/</ulink>. -</para></answer></qandaentry> - +</qandadiv> |